University of California Santa Cruz Creation And

Home , Coded wire tag, Natal homing, Salmon run

UNIVERSITY OF CALIFORNIA SANTA CRUZ

CREATION AND UTILIZATION OF NOVEL GENETIC METHODS FOR STUDYING AND IMPROVING MANAGEMENT OF CHINOOK SALMON POPULATIONS A dissertation submitted in partial satisfaction of the requirements for the degree of

DOCTOR OF PHILOSOPHY

OCEAN SCIENCES

Anthony J. Clemento

December 2013

The Dissertation of Anthony J. Clemento is approved:

Dr. John Carlos Garza, Chair

Dr. Jonathan Zehr

Dr. Grant Pogson

Dr. Eric Anderson

Dean Tyrus Miller Vice Provost and Dean of Graduate Studies Copyright c by

Anthony J. Clemento

2013 Table of Contents

List of Figures vi

List of Tables viii

Abstract xi

Dedication xiv

Acknowledgments xv

Introduction 1

1 Discovery and characterization of single nucleotide polymorphisms in Chinook salmon, Oncorhynchus tshawytscha 10 1.1 Abstract ...... 10 1.2 Introduction ...... 11 1.3 Methods ...... 16 1.3.1 Primer Design and PCR ...... 16 1.3.2 Sequencing and SNP Assay Development ...... 17 1.4 Results ...... 20 1.5 Discussion ...... 41

2 Evaluation of a SNP baseline for genetic stock identiﬁcation of Chi- nook salmon (Oncorhynchus tshawytscha) in the California Current Large Marine Ecosystem 48 2.1 Abstract ...... 48 2.2 Acknowledgments ...... 50 2.3 Introduction ...... 51 2.4 Methods ...... 55 2.4.1 Baseline Populations ...... 55 2.4.2 Markers and Genotyping ...... 60 2.4.3 Marker Selection ...... 61

iii 2.4.4 Population Genetics Analyses ...... 66 2.4.5 Power Analyses ...... 67 2.4.6 Mixed Fishery Samples ...... 69 2.5 Results ...... 71 2.5.1 Genotyping and Basic Population Genetics ...... 71 2.5.2 Assignment and Mixture Estimation Accuracy ...... 75 2.5.3 Fishery Sample ...... 76 2.6 Discussion ...... 79 2.6.1 Methodological Considerations ...... 80 2.6.2 Implications for Management ...... 84 2.7 Conclusions ...... 87

3 Large-scale genetic tagging experiment in a hatchery population of Chinook salmon (Oncorhynchus tshawytscha) allows for pedigree-based inference 88 3.1 Introduction ...... 88 3.2 Methods ...... 96 3.2.1 Study Site ...... 96 3.2.2 Hatchery Sampling ...... 98 3.2.3 DNA Extraction and Genotyping ...... 100 3.2.4 Population Genetic Analyses ...... 100 3.2.5 Pedigree Reconstruction ...... 101 3.2.6 Age Structure, Reproductive Success and Length-at-spawning . . 102 3.2.7 Relatedness ...... 104 3.2.8 Fishery Samples ...... 105 3.3 Results ...... 106 3.3.1 Population Genetic Parameters ...... 106 3.3.2 Hatchery Pedigree Reconstruction ...... 107 3.3.3 Age Structure ...... 111 3.3.4 Variance in Family Size and Reproductive Success ...... 113 3.3.5 Heritability of Length-at-spawning ...... 115 3.3.6 Relatedness ...... 119 3.3.7 Fishery Samples ...... 123 3.4 Discussion ...... 128 3.4.1 Technical Issues ...... 129 3.4.2 Parentage Assignments ...... 130 3.4.3 Heritability of Length-at-maturity ...... 131 3.4.4 Age Structure of Returning Adults and Spawning Broodstock . . 133 3.4.5 Inbreeding and Reproductive Success ...... 135 3.4.6 Fishery Assignments ...... 137 3.5 Conclusions ...... 139

Conclusions and Future Directions 140

iv References 144

v List of Figures

2.1 Unrooted neighbor-joining tree based on chord distances of 67 Chinook salmon populations from California to Alaska in the GSI baseline (see Table 2.1 for population details). Dashed lines indicate the position of populations which fall at tree junctions or have very short branch lengths. Sinona Creek and the coho salmon were omitted for missing data. . . . 74 2.2 Estimates of mixing proportions from cross-validation over gene copies (CV-GC) and K-Fold simulations for the eight most abundant reporting units in California Chinook salmon fisheries. The x-axis gives the true proportion of fish from each reporting unit, and the y-axis gives the estimated proportion. The dashed line is the y=x line. Shaded regions give the range between the 5% and 95% quantiles of estimates that would be achieved with perfect assignment of fish to reporting unit; i.e., they represent the uncertainty due to the fact that fishery proportions are estimated with a finite sample (in our simulations, a sample of 200 fish). The 5% and 95% quantiles of the estimates using genetic data from the CV-GC and the K-Fold methods are shown with vertical line segments and open diamonds, respectively. The mean over 20,000 CV-GC simulation replicates and 1,000 K-Fold replicates are given by filled circles and open triangles, respectively. These points fall along the dotted line when the estimator is unbiased...... 77

3.1 Age structure of returning adults (male and female) for two cohorts (2006 and 2007) from the Feather River Hatchery, CA. Numbers in parentheses indicate the total number of fish in each category, while white bars denote two-year olds, grey bars three-year olds and black bars four-year old fish. 112 3.2 Age structure of spawning adults (male and female) for two years of spawner broodstock from the Feather River Hatchery, CA. Numbers in parentheses indicate the total number of fish in each category, while white bars denote two-year olds, grey bars three-year olds and black bars four- year old fish...... 114

vi 3.3 Number of offspring that returned to the hatchery for females (white bars), males (grey bars) and mated pairs (dark bars) over all study years. The similarity over comparisons is expected as generally one male is spawned with one female at the hatchery...... 116 3.4 Number of offspring (full-siblings) that returned to the hatchery for parents spawned in each study year, 2006-2009. Note that offspring of 2009 spawners are under-represented as sampling permitted assignment of only two-year old fish...... 117 3.5 Relationship between the length of a mother and the number of her offspring that returned to the hatchery as adults at ages two, three or four. The size of full-sibling families here ranges from one to thirteen. . . . . 118 3.6 Linear regression of parental length on the length of their 3-year old adult offspring. Independent comparisons were made for: mean parent length and all offspring, male offspring, and female offspring, as well as, fathers and male offspring and mothers and female offspring...... 120 3.7 Distribution of the relatedness coeffcient (Rxy; Queller and Goodnight 1989) between all possible pairs of individuals in each collection of spawning broodstock and over all samples. Values are normally distributed, so the range, mean, standard deviation (Std. Dev.) and skew are reported. 122 3.8 Mated pairs were recorded at the FRH for spring-run spawners from 2006-2009. Parentage assignment allowed for the comparison of the distribution of relatedness (Rxy) among pairs that successfully had offspring return to the hatchery as adults (left side) and those that did not (right side). Again, values were normally distributed, and the range, mean, standard deviation (Std. Dev.) and skew are reported...... 124 3.9 Linear regression of the degree of relatedness between a parent pair (as estimated by Rxy) and the number of offspring that returned to the hatchery in subsequent years. This includes Rxy values for parents that had no offspring return...... 125

vii List of Tables

1.1 Summary of EST sequencing effort to identify genetic variation in populations of Chinook salmon (O. tshawytscha) from the west coast of North America. The weighted estimates account for unobserved variation in consensus sequence derived from less than 24 individuals...... 21 1.2 Description of the 117 SNP assays developed in this project with the target polymorphism, primer and probe sequences, length of the consensus sequence in base pairs (bp), and GenBank (dbGSS) and NCBI (dbSNP) accession numbers indicated...... 24 1.3 Summary statistics for 117 SNP loci in five Chinook salmon populations. N is the number of individuals genotyped. HE is expected (unbiased) heterozygosity and HO is observed heterozygosity. FST is over all five populations. AF is the observed frequency of the minor allele from the Feather River stock in each population. Asterisks (*) indicate significant (p<0.001) deviations from Hardy-Weinberg equilibrium...... 31 1.4 Preliminary BLAST results (BLAST hit and e-value) and annotation of the target SNP for the loci described here (Reference 1) and for an additional 24 loci (References 2, 3, 4 and unpublished) that are part of the final genotyping panel described in Chapter 2. Also included is whether the variation is present in an intron or exon and its location with respect to the described gene, either in coding sequence (CDS) or untranslated regions (UTR). No translation (n.t.) was available for 10 loci. For CDS exons, a single amino acid is indicated for synonymous substitutions, while both amino acids are included for non-synonymous substitutions. Reference codes are as follows: 1. Clemento et al. 2011; 2. Smith et al. 2005a; 3. Campbell and Narum 2008; 4. Smith et al. 2005b. 37

viii 2.1 Populations and reporting groups in the single-nucleotide polymorphism baseline for genetic stock identification of Chinook salmon from the West Coast of North America. Shown are the names used on the phylogeographic tree (Figure 2.1), the total number of individuals sampled (n), the number used in the training set (nt), estimates of unbiased expected (Exp.) and observed (Obs.) heterozygosity (Hz), and the mean number of alleles (A); also shown are the proportion of individuals that self-assign (Assign.) to the population (pop.) from which they were sampled and the proportion that self-assign to the correct reporting (rep.) group, as well as the mean FST for each population within and between reporting groups. Note that mean summary values shown were calculated excluding the coho salmon sample...... 57 2.2 List of the 96 single nucleotide polymorphism loci used to construct the baseline for genetic stock identification of Chinook salmon from the West Coast of North America, including dbSNP accession numbers (at the NCBI on-line repository for short genetic variations) and source reference (SR) where available: 1. Clemento et al. 2011; 2. Smith et al. 2005a; 3. Campbell and Narum 2008; 4. Smith et al. 2005b.; 5. Narum et al. 2008 64 2.3 Genetic stock identification (GSI) results of assigning 2010 California Chinook salmon fishery samples to their source populations using the single nucleotide polymorphism baseline, and concordance with coded- wire tag (CWT) recoveries...... 78

3.1 Summary of sampling and genotyping effort and success at the Feather River Hatchery, Oroville, CA. Included are the year of spawning, the hatchery designation of the run (Spring or Fall), the number of individuals genotyped (males and females), the number of individuals excluded for missing genotypes at more than 10 loci, and the number of individuals spawned, as reported by the hatchery. Population genetic parameters of unbiased heterozygosity (Hz), observed heterozygosity (Ho), the inbreeding coefficient (FIS; values with an asterisk are significantly different from zero, p<0.05, 1000 permutations) and the mean individual relatedness coefficient (Rxy; see text) were calculated for each genotyped broodstock year and spawn run. Spring origin fish among Fall-run spawners were identified using coded-wire tag data and included as putative offspring during parentage analysis...... 99

ix 3.2 Summary of offspring (offs.) recoveries using PBT from four spawn years of Chinook salmon from the Feather River Hatchery, CA. Reported are the number (n) of males and females with recovered offspring, the proportion of parent pairs included in the parent database (see text), the number of mated pairs recorded at the hatchery and the number of those matings confirmed by parentage analysis. A scaled estimate of the number of expected recoveries had all parents been included in the database is also shown...... 109 3.3 Heritability (h2) of length-at-maturity estimated as the slope of the length- length regression line between different comparisons of parents and offspring (Figure 3.6). Mean is the average length of the parents. The regression goodness of fit (R2) and standard error (SE) are also reported. 119 3.4 Summary of sampling effort and genotyping success in samples collected from mixed-stock ocean fisheries. Individuals with missing data at more than 10 loci were excluded. The age composition (comp.) of individuals in each sample was determined by assigning them to parents from the Feather River Hatchery (FRH). The number of the reported assignments (Assmts.) that were to fall-run parents sampled only in 2008 is also shown, as well as the age-class represented by those offspring (offs.) recoveries. For the 2010 samples collected at California ports, coded-wire tag information was available for comparison to genetic recaptures. . . . 127

x Abstract

Creation and utilization of novel genetic methods for studying and improving

management of Chinook salmon populations

Anthony J. Clemento

As a major component of fisheries in the northern Pacific Ocean, Chinook salmon (On- corhynchus tshawytscha) are of significant management concern. Their anadromous life history, in which adult fish migrate to their natal streams, leads to populations

(stocks) that are genetically distinguishable and, ideally, would be managed indepen- dently. Many of these stocks, particularly at the southern end of the species’ range, have experienced serious declines, which has motivated widespread hatchery production and supplementation. The physical coded-wire tagging (CWT) program currently used to track hatchery fish, and ultimately to supply information for cohort-based fishery harvest models, is increasingly ineffective and can no longer sustain the data demands of fishery managers and scientists. Also, current genetic tools utilizing microsatellite markers do not scale well to the enormous number of fish that need to be analysed, have error rates that are too high for individual- and pedigree-based methods, and genotype inconsistently across laboratories, creating an impediment to interjurisdic- tional collaboration. However, the next generation of genetic markers, single nucleotide polymorphisms (SNPs), do have low enough error rates and are amenable to the high- throughput genotyping required for ocean fishery stock identification and large-scale

xi tagging of hatchery fish via pedigree reconstruction. Here we describe the successful identification of 117 novel SNP loci using genomic data from a sister salmonid taxon and demonstrate their substantial power for discriminating five major stocks of salmon from the three largest basins on the Pacific coast of North America. We then assemble a panel of 96 SNP loci and genotype over 8000 individuals from 69 distinct populations for construction of a baseline for genetic stock identification (GSI) and show that it has, effectively, near-maximum power for discriminating most Chinook salmon stocks captured in mixed-stock fisheries off the coasts of California and Oregon. This baseline is used to confidently assign over 2000 ocean-caught Chinook to their source population and demonstrate over 99% concordance between the GSI assignments and identifications from CWTs recovered from these fish. The same panel of SNPs is also used to implement a large parentage-based tagging (PBT) experiment at one of the most productive hatcheries in the Central Valley of California. PBT involves genotyping reproducing adults and using their genotypes as intergenerational genetic tags that are recovered through parentage inference with their progeny. By genotyping over 12,000 individuals from six complete brood years, we show that the large number of resulting pedigrees effectively provide the same age and stock information as traditional CWTs, but also can be used to inform hatchery breeding practices, estimate the heritability of physical traits and eventually can serve as the basis for detailed linkage maps and associated mapping of quantitative trait loci. The genetic resources developed here are a substantial improvement over current methods and are fundamentally changing the way salmon populations are studied, monitored and managed.

xii xiii To Danielle and Sylvia, my family and friends, without whom this would never

have happened.

xiv Acknowledgments

While many people deserve recognition for their contributions towards the successful completion of this dissertation, none more so than my advisor John Carlos Garza. When

I was just another voice on the telephone asking for help, Carlos gave me the chance to come to Santa Cruz and work with the most amazing group of people and scientists that I have ever encountered. His unwavering conﬁdence in my abilities, and unending tolerance of my moods, has made me a better scientist, collaborator and researcher. He has brought me into this new family of West Coast salmon geneticists and made me feel right at home.

I would also like to express my gratitude to my committee: Jon Zehr, Grant Pogson and Eric Anderson for their patience while I found my way to the end of this journey.

I greatly appreciate the time they spent reading, editing and discussing my work and I give them full credit for being great role models of what a scientist should be. Special recognition to Eric, who has taught me to love the command line and has spent count- less hours passing on his extreme knowledge of all things computer.

Many thanks to the entire Molecular Ecology team at the Santa Cruz Lab. I owe my lab skills to the patient tutelage of Libby Gilbert-Horvath and would not have enjoyed using them without the companionship of the original lab crew: Big (Aguilar) and Lit- tle (Martinez) Andy, Scott Blankenship and Cheryl Dean. I have also learned so much from the other graduate students and post-docs who have cruised through the lab, all of whom I now consider life-long friends: Hilary Starks, Vicky Pritchard, Eric Crandall,

xv and Dan Barshis. I need to send a special thanks to Devon Pearse for sharing an oﬃce those many years and sharing his insight and experience almost every day. On a level of her own, Alicia Abad´ıa-Cardosohas been the best co-pilot ever as we have been on this graduate student journey together. I very much look forward to continued work with Martha, Diana, Cassie and Vanessa - you are all truly excellent.

I would be remiss if I did not mention my wife Danielle and my daughter Sylvia; their patience and tolerance during some of the more trying times was admirable and often it was only their love and caring that made everything alright. When there seemed to be no end in sight, they never lost faith in me and their conﬁdence in my ability to git-er-done was unwavering. And ﬁnally, I thank my parents, who gave me every opportunity to succeed, and I can only hope that I have exceeded their expectations.

The text of this dissertation includes reprints of the following previously published and submitted material:

Chapter 1: Clemento, A.J., A. Abad´ıa-Cardoso,H.A. Starks, and J.C. Garza. 2011.

Discovery and characterization of single nucleotide polymorphisms in Chinook salmon,

Oncorhynchus tshawytscha Molecular Ecology Resources 11(Suppl. 1):50-66.

I performed the majority of the DNA sequencing, data analysis and manuscript prepa-

ration for this project. H.A. Starks and A. Abad´ıa-Cardosoassisted with some of the

laboratory work and provided valuable suggestions for analyses and presentation of re-

sults. J.C. Garza (committee chair) directed and supervised the method development

and provided assistance with the writing and editing of the ﬁnal manuscript.

xvi Chapter 2: Clemento, A.J., E.D. Crandall, J.C. Garza and E.C. Anderson. 2014.

Evaluation of a SNP baseline for genetic stock identiﬁcation of Chinook salmon (On- corhynchus tshawytscha) in the California Current Large Marine Ecosystem Fishery

Bulletin, In Review.

For this project, I performed all of the genotyping of baseline populations and carried out the majority of analyses and manuscript preparation. E.D. Crandall coordinated genotyping of the ocean ﬁshery samples and compared results to coded-wire tag data.

E.C. Anderson directed the power analyses for marker selection and evaluation of the

ﬁnal genotyping panel for population discrimination. Both E.C. Anderson (committee member) and J.C. Garza (committee chair) directed and supervised the method development and provided assistance with the writing and editing of the ﬁnal manuscript.

This research was funded in part by a University of California Marine Council, Coastal

Environmental Quality Initiative (CEQI) grant (2006-2007) and a California Sea Grant/

CALFED Science Fellows Program grant (R/SF-24, 2007-2011).

xvii Introduction

The use of genetics to study and monitor populations is now ubiquitous as molecular markers have been discovered and made available for a wide range of species.

While the concept of genetic research and monitoring can be interpreted broadly, most implementations utilize molecular data to discern population structure, study evolutionary and ecological processes, or measure population genetic parameters over time

(Schwartz et al. 2006). Genetic markers can also be employed in lieu of physical tags for mark-recapture experiments (Palsbøl 1999) and as a tool for performing parentage assignment (Blouin 2003), which can yield insights into mating systems (Pearse 2001) and elucidate population dynamics (Hauser and Carvalho 2008). Pedigrees resulting from parentage assignment can also be used to estimate heritability and map genes involved in the inheritance of physical traits (e.g. size, growth, reproduction, migration) to their chromosomal locations (Fisher 1918, Lynch and Walsh 1998, Wu et al. 2007).

Such data can provide a predictive framework for assessing the eﬀects of environmental change on populations or the impacts of diﬀerent management and conservation actions on captive or wild populations.

As a keystone species in the marine, terrestrial and freshwater ecosystems of

1 the West Coast of North America (Willson and Halupka 1995, Cederholm et al. 1999,

Helﬁeld and Naiman 2006), Chinook salmon (Oncorhynchus tshawytscha) provide an excellent system for ecological and population genetic investigations. Chinook salmon are anadromous, a life-history strategy characterized by adult migrations from the ocean to spawn in their natal stream, followed by subsequent juvenile migrations back out to the ocean. This spawning site ﬁdelity creates genetically distinct populations and can provide an opportunity for local adaptation (Utter 1989; Taylor 1991). Chinook salmon exhibit broad variation in the duration of their juvenile freshwater residency, timing of adult spawning migrations and patterns of reproductive maturity (Taylor 1990, Groot and Margolis 1991). Two of the most common reproductive ecotypes described in the species, “spring-run” and “fall-run”, often inhabit the same river systems. Spring-run

ﬁsh are sexually immature as adult migrants, holding in deep river pools far up river drainages until they mature and spawn in fall and winter months; for fall-run ﬁsh sexual maturation is coincident with upstream migration in the Fall and consequent spawning in lower river reaches (Quinn 2005). Chinook populations are distributed around the

Paciﬁc Rim from the Central Coast of California in the east to Japan and coastal

Russia in the west. They are the target of highly valuable commercial and recreational

ﬁsheries throughout the northern Paciﬁc Ocean and continue to be a primary source of sustenance for Native American peoples.

Salmon, as with many other marine ﬁsh species (i.e. sardines and anchovies), naturally experience high variability in abundance. In the North Paciﬁc, salmon abundance has been shown to be tied to naturally occurring climate oscillations on decadal

2 timescales (Mantua et al. 1997). The predominant oceanographic features of the north- eastern Pacific are the eastern edges of the Alaskan subpolar gyre and the north Pacific subtropical gyre, which are alternately affected in the different phases of the Pacific

Decadal Oscillation (PDO; Hare and Mantua 2000, Mantua and Hare 2002). Likely driven by changes in primary production associated with sea-surface temperature and upwelling (Cole 2000, Hinke et al. 2005), salmon fisheries in Alaska and along the Pa- cific Northwest coast have alternately experienced depressions corresponding with PDO cycles (Kruse 1998, Hare et al. 1999). While fishery scientists can utilize this type of oceanographic information for increased ecosystem-based management (Field and Fran- cis 2006), salmon still face challenges not only in the marine environment but also in rivers and streams (Bisbal and McConnaha 1998).

Degradation of riverine spawning habitat, diversion of fresh water for human use, over-ﬁshing, hatchery domestication selection, and highly variable ocean conditions have all been implicated in the recent declines of Chinook salmon populations in the southern portion of the species’ range (Lindley et al. 2009). As a consequence, many

Chinook salmon populations in the contiguous United States are now listed as threatened or endangered under the federal Endangered Species Act (Myers et al. 1998). In order to mitigate for the multiple impacts threatening Chinook salmon populations, state and federal agencies now produce millions of fish annually in hatcheries. These hatchery fish, intended to reduce variability in ocean abundance, provide fishing opportunities, and satisfy Native American treaty obligations, comingle with wild fish in the ocean and can compose the majority of the catch in certain times and places (Beamish

3 et al. 1997). However, the ecological consequences of releasing large numbers of hatchery fish are poorly understood (Levin et al. 2001), and may be severely compromising efforts to preserve wild populations (Hilborn 2011). Some natural populations may now be composed primarily of hatchery individuals or their offspring (Barnett-Johnson

2007).

The Central Valley of California was once the second largest source of Chinook salmon on the U.S. West Coast (after the Columbia River), despite being the southern- most drainage to support the species. Dominated by the Sacramento River to the north and the San Joaquin River to the south, the Central Valley historically maintained wild populations that numbered in the millions (Yoshiyama et al. 2001). As in most river systems that support the species, Chinook salmon from the Central Valley display a wide variety of life-history strategies, varying in the timing of migrations and sexual maturation (Yoshiyama et al. 1998). However, a majority of the historical spawning habitat for salmon has been eliminated as rivers are engineered for ﬂood control and water is appropriated for agriculture, domestic water supplies, and hydroelectric production (Fisher 1994). Because spring-run Chinook salmon, which migrate upstream months before the fall-run form, generally penetrate further up into watersheds, the many large dams on Central Valley rivers disproportionately eliminated their primary spawning and holding habitats. Spring-run Chinook, which historically were more numerous than the fall-run in ocean ﬁsheries (Yoshiyama et al. 2001), have experienced severe declines in California and are now listed as threatened under the California state and the federal endangered species acts. Six hatcheries now produce the majority of

4 Chinook (spring- and fall-run) that return to the Central Valley (Fisher 1994).

Ensuring sustainability and the persistence of salmon populations while providing fishing opportunities can be a complex task. Underestimation of the contribution from specific stocks can have serious conservation implications (e.g. overfishing and/or extinction of wild stocks), while overestimation can leave the resource underexploited, potentially costing the fishing industry and coastal communities millions of dollars (Michael 2010). Management of Pacific Ocean salmon fisheries off North America can be roughly divided into three regions: California and Oregon fisheries are managed by the Pacific Fishery Management Council (PFMC); fisheries in Washington, British

Columbia, Canada and southeast Alaska are subject to the international Paciﬁc Salmon

Treaty, reported to and regulated by the Pacific Salmon Commission (PSC); and fisheries further north and west in Alaska are managed by the state, with salmon by-catch under the purview of the North Pacific Fishery Management Council. The primary method of assessing fishery impacts is through cohort analysis models. These models attempt to account for fishery mortalities on groups of fish (primarily of hatchery origin) born in the same year (and therefore the same age) through time. However, the uncertainty in the models coupled with the difficulty of estimating fishery impacts on highly age-structured populations has left scientists and managers in need of better data (Hankin 2005). Currently, the primary source of information for cohort analysis models comes from coded wire tags. Management of Chinook salmon fisheries in the eastern Pacific Ocean depends on an elaborate marking and coded-wire tagging program, implemented and monitored by State and Federal agencies. The primary focus

5 of this monitoring program is the millions of fish produced annually in hatcheries along the West Coast of North America. Data extrapolated from the program are used to parameterize stock-specific forecasting models and to estimate ocean abundance indices, which are then used to set fishing areas and seasons, determine quotas and legal gear and establish catch limits and size restrictions (Hyun, 2012). The accuracy of these models and the resulting abundance estimates are highly dependent on the quantity and quality of data input.

Genetic methods have long been used to study various aspects of salmon biology and ecology. The earliest genetic analyses of salmon utilized electrophoretically detectable protein polymorphisms known as allozymes, which were suﬃcient for discriminating populations at a relatively coarse geographic scale (Milner et al. 1985; Tessier et al. 1995; Allendorf and Seeb, 2000). With the introduction of polymerase chain reaction (PCR) and development of modern genetic techniques, a variety of new marker types became available to salmon scientists, including mitochondrial markers (mtDNA), ampliﬁed fragment length polymorphisms (AFLPs), minisatellites, and microsatellites

(Beacham et al. 1996; Smith et al. 2001; Flannery et al. 2007; Clemento et al. 2009).

For almost two decades, microsatellites have been employed in studies of population structure, behavioral ecology, and pedigree relationships, as well as for individual and genetic stock identiﬁcation, because of their extensive polymorphism (Banks et al. 2000;

Smith et al. 2005b/c; Seamons et al. 2004; Pearse et al. 2007; Smith et al. 2007). While their high variability provides suﬃcient statistical power for many population genetic applications, microsatellites can have high genotyping error and mutation rates. In ad-

6 dition, combining microsatellite data generated in different laboratories, or on different instrument platforms in the same laboratory, can require an onerous standardization procedure to account for subtle differences in electrophoretic conditions and resulting instrument output. This standardization process can add significant time and expense to multilateral database construction and collaborative research (Seeb et al. 2007).

As genomic resources for salmonids have expanded, single nucleotide polymorphism (SNP) markers have become an increasingly common choice for population genetic studies (Morin et al. 2004). A SNP is variation at a single DNA base at a known location in the genome. SNPs are abundant and can be found in both coding regions, where they may be targeted by selection, and in non-coding regions, where they are often assumed to not be the direct targets of natural selection (Vignal et al. 2002, Nosil et al. 2009). Since SNPs are generally bi-allelic, comparable power to highly polymorphic microsatellite markers is attained by using larger numbers of loci (Anderson and

Garza 2006, Narum et al. 2008). At the same time, new technologies have yielded platforms (i.e. nanofluidics, microarrays) for efficient, high-throughput genotyping at large numbers of SNP markers. Data generation on these platforms requires significantly less time and money and resulting genotypes are subject to much lower error rates than microsatellites. While SNP development for Chinook salmon began around the middle of the last decade (Smith et al. 2005a, 2005b, 2006; Campbell and Narum 2008), only about 30 markers were available at the outset of the research described here. Discovery of larger numbers of SNPs (described in Chapter 1) is the first step towards implementing new SNP-based methods for genetic stock identification (GSI; Chapter 2) and

7 parentage-based tagging (PBT; Chapter 3) of Chinook salmon.

In Chapter 1, I further motivate the need for SNP development in Chinook salmon and describe the methodology we used to discover 117 novel SNP markers. De- spite the broad importance of the species, very few genomic resources are available, so the utility of expressed sequence tag (EST) data from a more well-studied, somewhat closely-related, species was explored, steelhead trout (Oncorhynchus mykiss), for primer design. Although ESTs are by deﬁnition in coding regions, it can be expected that the resulting genomic DNA sequences will encompass introns, where SNPs may also be observed. I hypothesize that employing a balanced ascertainment panel for sequencing, with representatives drawn from a broad geographic range, will yield SNPs with increased power for population discrimination. Furthermore, implementing strict criteria about observed genotypes in the sequencing data should high-grade for SNPs with suﬃcient allele frequencies for GSI and PBT applications.

In Chapter 2, I describe the development of a coastwide genetic database for identifying Chinook salmon caught in ﬁsheries in the California Current Large Marine

Ecosystem. It is hypothesized that a single panel of 96 SNPs can be sufficient for providing the stock of origin for fish captured in large mixed-stock ocean fisheries. I explain the procedures used to select SNPs for inclusion in the panel and evaluate the power of the new baseline for genetic stock identification using valid statistical methods. The markers designed using the balanced ascertainment strategy described in Chapter 1 can be expected to be particularly effective for GSI, even if there are not large allele frequency differences between populations. I also demonstrate that inference

8 from the genetic data is comparable to that generated by physical tags for management applications.

In Chapter 3, I hypothesize that the same panel of SNP markers employed in Chapter 2 will be equally effective for intergenerational genetic tagging of a large hatchery population from the Feather River Hatchery, CA. The parentage-based tagging technique described here is fundamentally different from GSI; rather than assigning fish to their most likely management unit using allele frequencies, individuals are specifically identified by inferring parentage. The method is expected to be sufficiently powerful that the parent pairs assembled in pedigrees using the genetic data will match those recorded during spawning at the hatchery. Additionally, I demonstrate the utility of knowing large numbers of pedigrees for estimating the impacts of artificial propagation in the hatchery. Pedigrees can be used not only to assess important population genetic parameters (i.e. inbreeding) in unprecedented detail but also to estimate the heritability of observed life-history traits. It is likely that the tools developed and described here will substantially change the way Chinook salmon are managed, both in ocean fisheries and at hatcheries.

9 Chapter 1

Discovery and characterization of single nucleotide polymorphisms in Chinook salmon, Oncorhynchus tshawytscha1

1.1 Abstract

Molecular population genetics of non-model organisms has been dominated by the use of microsatellite loci over the last two decades. The availability of extensive genomic resources for many species is contributing to a transition to the use of single nucleotide polymorphisms (SNPs) for the study of many natural populations. Here we describe the discovery of a large number of SNPs in Chinook salmon, one of the worlds most important ﬁshery species, through large-scale Sanger sequencing of expressed se-

1published: Clemento, A.J., A. Abad´ıa-Cardoso, H.A. Starks, and J.C. Garza. 2011. Molecular Ecology Resources 11(Suppl. 1):50-66.

10 quence tag (EST) regions. More than 3MB of sequence was collected in a survey of variation in almost 132KB of unique genic regions, from 225 separate ESTs, in a diverse ascertainment panel of 24 salmon. This survey yielded 117 TaqMan (5’ nuclease) assays, almost all from separate EST regions, which were validated in population samples from ﬁve major stocks of salmon from the three largest basins on the Paciﬁc coast of the coterminous United States: the Sacramento, Klamath and Columbia Rivers. The proportion of these loci that was variable in each of these stocks ranged from 86.3 to

90.6% and the mean minor allele frequency ranged from 0.194 to 0.236. There was substantial differentiation between populations with these markers, with a mean FST estimate of 0.107, and values for individual loci ranging from 0 to 0.592. This substantial polymorphism and population-specific differentiation indicates that these markers will be broadly useful, including for both pedigree reconstruction and genetic stock identification applications.

1.2 Introduction

Chinook salmon (Oncorhynchus tshawytscha) is the largest species of Pacific salmonid and one of the worlds most commercially and recreationally valuable fishery species. Chinook salmon are anadromous, meaning that they hatch in rivers and streams, migrate to the ocean during either the first or second year of life, and then typically return to their natal stream to spawn. This homing creates geographic population structure and facilitates the potential local adaptation of populations and larger groups.

11 In the marine environment, stocks from different rivers, hatcheries and ecotypes, as well as fish of different ages, commingle, making it difficult to quantify catch composition or avoid stocks with depressed abundance in ocean fisheries. Degradation of riverine spawning habitat, diversion of fresh water for human use, over-fishing, hatchery domestication selection, and highly variable ocean conditions have all been implicated in the recent declines of populations in the southern portion of the species range (Lindley et al. 2009). As a consequence, many Chinook salmon populations in the contiguous

United States are now listed as threatened or endangered under the federal Endangered

Species Act (Myers et al. 1998). Populations in California have seen particularly severe reductions over the last decade, culminating with complete closures of the commercial

ﬁshery oﬀ California and Oregon in 2008 and 2009 (Lindley et al. 2009).

Population genetics has played a prominent role in salmon research and management over the last several decades. However, the predominant type of molecular genetic marker used has varied substantially over time. Prior to and immediately following the introduction of the polymerase chain reaction (PCR), allozymes were the primary type of genetic marker available for ﬁsh biologists (Myers et al. 1998; Waples et al. 2004). Following the introduction of PCR, came many other marker types including mitochondrial (mtDNA), ampliﬁed fragment length polymorphisms (AFLPs), minisatellites, and microsatellites (Beacham et al. 1996; Smith et al. 2001; Schlotterer 2004;

Flannery et al. 2007; Clemento et al. 2009). Microsatellites, in particular, have been em-

ployed broadly in salmonids for studies of population structure, behavioral ecology, and

pedigree relationships, as well as for individual and genetic stock identiﬁcation, because

12 of their extensive polymorphism (Banks et al. 2000; Smith et al. 2005b/c; Seamons et al. 2004; Pearse et al. 2007; Smith et al. 2007). This variation provides substantial statistical power for many population genetic applications, but related to this is that microsatellites can have high genotyping error and mutation rates. In addition, combining microsatellite data generated in different laboratories, or on different instrument platforms in the same laboratory, may require a non-trivial standardization process to account for subtle but ubiquitous differences in electrophoretic conditions and resulting instrument output. This standardization process adds significant time and expense to multilateral database construction and collaborative research (Seeb et al. 2007).

More recently, single nucleotide polymorphism (SNP) markers have come to prominence (Morin et al. 2004). A SNP is a variation in the base present at a specific nucleotide site in the genome. SNPs are the most abundant polymorphism in vertebrate genomes, with a SNP present every 100-500bp on average (Vignal et al. 2002). They are common in both coding and non-coding regions of the genome and are typically biallelic, so analytic power similar to that provided by microsatellites is achieved by using larger numbers of loci (Anderson and Garza 2006; Narum et al. 2008; Glover et al. 2010). SNPs require substantially less laboratory staff time for allele calling and with the advent of new high-throughput genotyping technology, such as nanofluidics and spotted microarrays, data can be generated more quickly and at lower cost than for other marker types. Moreover, standardization requires only that laboratories agree to reporting standards and, ideally, that they use an identical, or overlapping, set of markers. While SNPs have seen extensive use in humans and model organisms, other

13 research communities have been slow to transition to SNP-based data collection, primarily because of the lack of genomic resources available for non-model species and the costs and eﬀort involved in marker development.

In recent years, SNP development for Pacific salmonids has begun (Smith et al. 2005a, 2005b, 2006; Aguilar and Garza 2008; Campbell and Narum 2008; Campbell et al. 2009; Abad´ıa-Cardoso 2011) and there are currently a handful of SNP assays available for all of the Pacific salmonid species. Nevertheless, many more are necessary for a number of the applications in which genetic markers are currently in use, including pedigree reconstruction, genetic stock identification, linkage map construction and QTL mapping. Moreover, since many of the existing assays were developed with specific applications in mind, they are frequently of limited utility in populations or phylogenetic lineages that were not part of the discovery process, due to ascertainment bias (Clark et al. 2005; Smith et al. 2007; Albrechtsen et al. 2010). The implementation of SNP- based methods, such as large-scale parentage inference (Garza and Anderson 2007), in

California and other marginal parts of the species range requires many additional SNP assays.

SNP discovery typically involves examination of DNA sequence data from multiple individuals at the same locus, or identiﬁcation of heterozygous nucleotide sites in a single individual. When only a small number of individuals from selected populations are used to discover or ascertain SNP variation, an ascertainment bias is introduced.

This bias results in the allele frequency spectrum being shifted upward, with an under- representation of rare SNPs, which leads to overestimates of genome-wide heterozygosity

14 and population diﬀerentiation (Clark et al. 2005; Smith et al. 2007; Albrechtsen et al.

2010). This ascertainment sampling bias also results in SNPs that are not as polymorphic in other parts of the species range. Clark et al. (2005) recommend the use of standardized ascertainment criteria and a large ascertainment sample of known origin that includes individuals from outside of the primary focus range to reduce these biases and provide marker loci with broad utility.

Despite the species’ importance, genetic resources for salmonids are still relatively scarce; there is not a complete genome sequence for any salmon species and there is not even a published linkage map for Chinook salmon. There is, however, a large library of expressed sequence tags (ESTs) from rainbow trout (Oncorhynchus mykiss), archived by the Gene Index Project (http://compbio.dfci.harvard.edu/tgi/). These ESTs include full or partial cDNA sequences that have been reverse transcribed from mRNA (Bouck and Vision 2007). Rainbow trout is in the same genus as Chinook salmon and previous work has shown that primers derived from O. mykiss can be successfully used to isolate

DNA fragments and discover SNPs in other Oncorhynchus species (Smith et al. 2005a).

Primers for conserved regions of known genes are also available in the literature (e.g.

Moran 2002).

Here we describe the discovery, design of molecular assays, and evaluation of

117 new SNP assays for Chinook salmon, more than doubling the number of published

SNP markers for use in the species. We sequenced genic regions in the Chinook genome from a geographically and phenotypically diverse ascertainment sample of 24 ﬁsh. We targeted 480 loci from ESTs of unknown function, as well as genes whose functions are

15 well described, and designed more than 150 5’ exonuclease (TaqMan) assays. Assays were tested and validated by genotyping 337 individuals from ﬁve major lineages of the species, from the three largest rivers on the west coast of the coterminous United

States (Sacramento, Klamath and Columbia Rivers) and the details of the resulting 117 validated assays are reported here.

1.3 Methods

1.3.1 Primer Design and PCR

Oligonucleotide primers were designed for 480 ESTs randomly selected from the O. mykiss Gene Index database of EST sequences. A secondary targeted gene approach was undertaken using primer information from 11 genes from published sources

(e.g. Moran 2002) or in GenBank. Primers were designed using Primer3 (v.0.4.0; Rozen and Skaletsky 2000) and targeted EST segments 400-500 bp in length, so that genomic

DNA fragments would generally be smaller than 1000bp, even if they contain introns.

These primers (sequences are available from the authors upon request) were then used to amplify genomic DNA from a geographically and phylogenetically diverse ascertainment sample of 24 Chinook salmon, including ﬁsh from California populations with which we are actively working (Sacramento: Feather River-Spring and Fall, n=12; Sacramento:

Butte Creek-Spring, n=2; Eel River, n=2; Klamath River, n=2), and also from elsewhere in the North American range of the species, including Washington (Columbia-Kalama

River-Spring, n=2) and Canada (Thompson River-Spius Creek, n=2; Nanaimo Creek,

16 n=2). For the California samples, DNA was extracted from dried caudal ﬁn clips using

DNeasy 96 kits on a BioRobot 3000 (Qiagen, Inc.). For the other samples, previously extracted and frozen DNA was provided by collaborators.

Polymerase chain reaction (PCR) was carried out in 15µL single-locus reactions using Applied Biosystems (ABI) reagents as follows: 1.5µL of 10X buﬀer, 0.9µL of

1.5mM MgCl2, 1µL of 2.5mM dNTPs, 1µL of 5mM primers (forward and reverse), 6.6µL of deionized water, 0.05µL of AmpliTaq 5U/µL DNA polymerase, and 4µL of genomic

DNA. The thermal cycling routine employed was a modiﬁed step-down protocol with an initial denaturation of 95◦C for 5 min, followed by 95◦C for 3 min, 63◦C for 2 min,

72◦C for 1 min, repeated 13 times with a 1◦C decrease in anneal temperature (63-50◦C) each cycle, then 9 cycles of 95◦C for 30 s, 51◦C for 30 s, 72◦C for 1 min, and 11 cycles of

95◦C for 30 s, 51◦C for 30 s, 72◦C for 1 min (+10 s/cycle), with a ﬁnal 5 min extension at 72◦C. PCR products were visualized on 2% agarose gels by electrophoresis.

1.3.2 Sequencing and SNP Assay Development

If a locus showed a single band on the agarose gel for most individuals, PCR products from all 24 individuals were then sequenced, even for samples from which no band was visible, in an eﬀort to preserve sample sizes. An EXO-SAP clean-up was performed prior to sequencing; 5µL of PCR product, 0.15µL of Exonuclease I (20U/µL),

1µL of shrimp alkaline phosphatase (1U/µL), 0.5µL of 10x buﬀer and 3.36L of deionized water were incubated at 37/µC for 60 min and then 80◦C for 20 min with a ﬁnal cool down to 4◦C. Cycle sequencing reactions employed the BigDye Terminator sequencing

17 kit (v. 3.1; Applied Biosystems, Inc.) with standard conditions. Sequencing reaction products were then puriﬁed using 6% Sephadex columns and sequenced on an ABI

3730 DNA Analyzer using standard conditions. Sequences were assembled into contigs

(24 individuals, forward and reverse sequences) and aligned with Sequencher 4.6 (Gene

Codes Corporation) using the Dirty Data algorithm with a Minimum Match Percentage of 85% and a Minimum Overlap of 20 bp. Potential polymorphisms were visually veriﬁed on the chromatograms.

Only sites for which both homozygote genotypes and the heterozygote genotype were observed were chosen for assay development, so as to minimize identification of sequencing artifacts as polymorphisms and to ensure that the resulting SNP assays would have suitable minor allele frequencies for our intended applications. If all observed variable sites were heterozygous, we assumed that this was likely a duplicated gene and it was excluded from further analyses. In consensus sequences with multiple candidate SNPs, the site with the highest minor allele frequency (MAF) in the sequences from the Feather River populations was selected. The location of the SNP in either exonic or intronic sequences was evaluated (see Table 1.4) but was not used as a criteria for selecting the target variation. The contig sequence information was then sent to ABI for design of 50 exonuclease (TaqMan) assays. Taqman assays use two sequence-specific unlabeled primers and two allele-specific fluorescently-labeled probes to directly distinguish nucleotide variants (SNPs) in the target genomic DNA sample.

These assays can be interrogated on a single locus, real time PCR instrument (e.g. ABI

7300 Real Time PCR System) or on a multiplex platform (e.g. Fluidigm BioMark/EP1

18 nanoﬂuidic arrays).

Each Taqman assay was then evaluated on population samples of salmon from the three largest basins on the West Coast of the United States, in order to validate assay performance, reﬁne allele frequency estimates and to evaluate the expected power of the markers for various applications. The ﬁve populations/stocks that were included are Feather River Spring-run, Butte Creek Spring-run, and Mokelumne River/Battle

Creek Fall-run from the California Central Valley, Klamath and Trinity River Fall-run from northern California, and the Kalama and Cowlitz Rivers Spring-run stocks from the Columbia River basin in Washington. The 337 individuals from these ﬁve populations/stocks were genotyped with all designed assays on Fluidigm 96.96 Dynamic

Arrays using the Fluidigm EP1 instrumentation and according to the manufacturers protocols. The Fluidigm system uses nano-ﬂuidic circuitry to simultaneously genotype up to 96 samples with 96 loci in tiny reaction chambers embedded on the arrays (see

Seeb et al. 2009 for a full description of the Fluidigm system methodology). Genotypes were called and the data compiled using the Fluidigm SNP Genotyping Analysis software. Each assay was assessed for plot quality and expected clustering patterns. The

MAF and expected (HE, unbiased) and observed (HO) heterozygosity were calculated for each population. The software package GENETIX (Belkhir et al. 1996-2004) was used to estimate global FST (theta) with the estimator of Weir and Cockerham (1984), as an indicator of the power of the locus for genetic stock identiﬁcation and related applications. Deviations from Hardy-Weinberg equilibrium proportions were evaluated with GENEPOP 4.0 (Rousset 2008).

19 1.4 Results

Of the 480 EST fragments targeted in the initial round of discovery, 244 yielded a single band when PCR products were electrophoresed in agarose and were further evaluated by sequencing; loci with multiple, weakly visible, or no PCR products were not considered further. Of the 244 loci that yielded PCR products, we successfully acquired sequence data for 225 EST fragments, with an average of 32 (of a maximum possible of 48) sequences per locus, when considering both forward and reverse strands

(Table 1.1). The total length of the consensus sequences generated was 131.3kb, with a mean consensus length of 554 bp per gene. Eighty-seven loci (38.7%) yielded fragments substantially larger than the target fragment (for 12 of these loci forward and reverse sequences did not overlap), indicating the presence of one or more introns. Of the 225

EST loci for which sequence data were obtained, 177 contained some variation. In total,

661 variable sites were observed (including substitutions and insertions/deletions) and, of these, 611 were observed nucleotide substitutions that are potential SNPs. Only two nucleotides were observed at all but two of the 611 sites with substitutions present, with three bases observed at the other two sites. Fifty insertion/deletion polymorphisms were also identiﬁed, as were ﬁfteen suspected microsatellites, but these were not considered targets for assay development at this time.

The mean density of observed mutations in the ∼131 kb of consensus EST sequence was 0.0046, or about one substitution every 215 bp. The mean length of fragments composing the consensus sequence was also weighted by the number of in-

20 Table 1.1: Summary of EST sequencing eﬀort to identify genetic variation in populations of Chinook salmon (O. tshawytscha) from the west coast of North America. The weighted estimates account for unobserved variation in consensus sequence derived from less than 24 individuals.

Total Mean per locus [range]

EST loci successfully sequenced 225 Base-pairs sequenced (all fragments) 3,024,916 12763.36 [382-45720] Length of consensus sequence (bp) 131,287 553.95 [99-1566] Weighted consensus (bp) 112,115 498.29 [72-1524] Number of observed substitutions 611 2.72 [0 - 17] Number of SNPs (all three genotypes observed) 228 1.01 [0 - 7] Loci with no variable sites 48 Insertions/deletions (indels) 50 Transitions (A-G or C-T) 319 Transversions (A-C or G-C or A-T or G-T) 290 Sites with 3 nucleotides observed 2 Possible duplicated genes 11 Total number of substitutions + indels 661 Density of substitutions in consensus sequence 0.0047 Density of substitutions in weighted consensus sequence 0.0054

21 dividuals for which each nucleotide was sequenced, so as to correct estimates of SNP density for undiscovered variation in the unsequenced individuals. This weighted consensus sequence length yielded a density estimate of 0.0054 or about one SNP every

183 bp. When only candidate SNPs (all three genotypes observed in sequences) are considered, density in the consensus sequence was 0.0017 (or about one SNP per 576 bp), whereas the weighted density was 0.0020 (or about one SNP per 492 bp).

Only nucleotide sites where all three genotypes were observed in the sequence data were considered as candidates for assay development. There were 228 of these putative SNPs present in the sequence data, with from one to seven present per gene, and sites in 112 genes also met the criteria for TaqMan assay design (SNP more than

40bp from either end of the sequence, with no additional variation or ambiguous sites within two bp of the target SNP). Fifteen of the original assays failed to produce reliable genotype data in the validation populations, which was defined as: no signal (all plots at the origin), a single cloud of plots with no distinct clusters, more than three clusters, or no heterozygote but both homozygote clusters within a population. For ten of the assays that failed, there were other variable sites in the genes in which they were located that met both the ascertainment and the assay design criteria. However, only five of these redesigned assays produced reliable genotypes. In addition, one of the assays that initially failed, produced reliable results with a manual assay redesign, for a grand total of 103 validated assays from the EST sequencing effort. In the small, secondary discovery effort, a total of 14 polymorphic sites, in 11 candidate genes, met the ascertainment and assay design criteria, and all yielded reliable genotype data. Multiple SNPs were

22 designed for the Aldolase and NAML genes. For the ﬁnal 113 gene regions that contain the 117 validated assays, consensus sequences that indicate all of the observed nucleotide variation from the ascertainment sample were compiled and submitted to Genbank dbGSS (Accession Nos. HR308668-HR308783), while targeted SNP loci were uploaded to the NCBI dbSNP database (Accession Nos. 275518685-275518802; Table 1.2).

23 Table 1.2: Description of the 117 SNP assays developed in this project with the target polymorphism, primer and probe sequences, length of the consensus sequence in base pairs (bp), and GenBank (dbGSS) and NCBI (dbSNP) accession numbers indicated.

Assay name Targets Primers (50-30) Probes (50-30) bp dbGSS dbSNP Ots 94857-232 T/C F: GGCACTCTCCCTGGCTAGA VIC: CAGGATAATAACAAACAAG 687 HR308668 275518685 R: CCCCATCACTTCTCTGGCTTTAAAT FAM: CAGGATAATAACGAACAAG Ots 94903-99 G/T F: CCGTCTGAGTAGGAGGATCAATACA VIC: CAAACCAGCAAACAT 314 HR308669 275518686 R: TTTGGATCCAGCTCTCCGTATAGA FAM: ACAAACCAGAAAACAT Ots 95442b-204 T/A F: GTCTCTCTCTCTTTGCATCATTACACT VIC: TGGTTCCCCAAATTT 256 HR308670 275518687 R: GGACTCTTGAGCTGTCTGGCTATAT FAM: TGATGGTTCCCCTAATTT Ots 96222-525 C/T F: GCTCTTGCCCATCTGTAGGAT VIC: TGTAGCTAATTTTAAGTTCTC 651 HR308671 275518688 R: GGCGCAACATATGTATTAAGCAACT FAM: AGCTAATTTTAAATTCTC Ots 96500-180 G/T F: GATCATGTCAGATAGGATGCTGAAAGT VIC: AAAACAAATCATTTTTCG 313 HR308672 275518689 R: CAGGTCTGGTCTACATCGAACAC FAM: AAAAACAAATAATTTTTCG Ots 96899-357 T/A F: TCTCCTGAACTAATTTAGACCTCTGAATGT VIC: CTGAATGTTTTTTTTAATCTTT 577 HR308673 275518690 R: CCTCATATTGCTTTCATCTGAAGAGAGA FAM: CTGAATGTTTTTTTTTATCTTT Ots 97077-179 G/T F: CCTGAACAAATACTTAACGCTCCAGTT VIC: TCACAAATGTATCCTAAAGC 288 HR308674 275518691 R: GTAATAATACTTCACACCATTGCCACTTC FAM: CACAAATGTATACTAAAGC Ots 97660-56 T/A F: TTCCCTAATCTGACGTACTACCAACT VIC: ACGAGACAGATATTC 455 HR308675 275518692 R: CGCCACTGACGTTCATTCCA FAM: ACGAGACTGATATTC Ots 98409-850 C/T F: CTGCGTTTCTGGAATGTTTTCAGT VIC: TTGTTCACGAACCTTG 1072 HR308676 275518693 24 R: CAAACCTGTTACTGGCCAAATGAAA FAM: TTGTTCACAAACCTTG Ots 98683-796 T/A F: GCAATGGCATGACAATGGAAGTC VIC: CTCAGCCCCTATATTACAA 895 HR308677 275518694 R: CACTGGCACTGGTGGAGATTA FAM: CTCAGCCCCTATTTTACAA Ots 99550-204 C/T F: TGACAGATTTCACCTTTAACTAGCTAAGC VIC: AAGGCTTTGGTTGTTTG 356 HR308678 275518695 R: GCAACCTCTTTCACACTTCAGTAAC FAM: AAGGCTTTGATTGTTTG Ots 100884-287 T/C F: CGGAAGACCAGATTCTCCAAGAGTA VIC: ATAGAACTACAATTCACATATAT 470 HR308679 275518696 R: CGACCAAGTAGCGGCACTT FAM: AACTACAATTCGCATATAT Ots 101119-381 T/C F: TTTTCTAGGACAGGTTGCTTGCA VIC: TGCCACATGATAATTGA 1122 HR308680 275518697 R: CCAGGTTTCTTTAGCCTACTTATTCTTTACA FAM: CCACATGGTAATTGA Ots 101554-407 C/G F: TGAAAGATATCAATTGTAGTAGTGGTGGTG VIC: ATGGAGGATTGTGGTTGT 417 HR308681 275518698 R: ACACGCCAGTCCACAAGT FAM: ATGGAGGATTCTGGTTGT Ots 101704-143 T/G F: ACTTCTTGAGCCAATCGGATGATG VIC: CTTAGACGTCAGAGGTC 580 HR308682 275518699 R: CCAGAGATAAACTAGTGGAGGAGATCA FAM: CTTAGACGTCCGAGGTC Ots 101770-82 C/A F: GCGACTTGACAACGAGGAGAA VIC: ACTTCCCGGAGCTGC 783 HR308683 275518700 R: CCCTCTTCATAACGTTACCAAACAC FAM: ACTTCCCTGAGCTGC Ots 102195-157 T/C F: TGGTCAGCGGTGTCTTTCAC VIC: TGCTGATTCAAGAAGAAGTA 543 HR308684 275518701 R: CCCCGCATCTGTCAAATGGAT FAM: CTGATTCAAGGAGAAGTA Ots 102213-210 A/G F: CATTCCATGACAATGATTGAAATCTAAAAACAC VIC: CTGTATACAGTAAGAGTATTAAT 1074 HR308685 275518702 R: GAGTATCTCAATTGCAACACTATGGTATGT FAM: ACAGTAAGAGCATTAAT Ots 102414-395 A/G F: GCCTACTGATAAATGTATGACAGTAATGGA VIC: CACATAGTGTAGCTTTACTAC 1030 HR308686 275518703 R: CAATAACAAACAAGCTAGGAACAAAAGTGT FAM: CACATAGTGTAGCTCTACTAC Ots 102420-494 T/G F: TGCCAACCTGGCCAGTTAC VIC: CATGTGAACAACAAGCG 739 HR308687 275518704 R: GCTTCCCTGCTTCCATGGT FAM: CATGTGAACACCAAGCG Continued on next page Table 1.2 – continued from previous page

Assay name Targets Primers (50-30) Probes (50-30) bp dbGSS dbSNP Ots 102457-132 A/G F: CCAGCAGAGACTGGGTTCAC VIC: CAATTGTGCGTTGCCCCA 734 HR308688 275518705 R: TTCCCTACCGGCGAAACC FAM: ATTGTGCGTCGCCCCA Ots 102801-308 C/A F: TGGGACAGAGGTGGGAATTGA VIC: AGGGACAGTTTCGCAGACG 670 HR308689 275518706 R: CCCAAAGATGCTTAACTGAAGATGTG FAM: AAGGGACAGTTTCTCAGACG Ots 102867-609 A/G F: CTCTGCCATTCATTTGGGCTTTG VIC: ACAGAGAGAAGTCCCAGGTG 796 HR308690 275518707 R: GTCTAAAGTGGTCCCCTTGGAT FAM: AGAGAGAAGCCCCAGGTG Ots 103041-52 G/A F: ACCACCCACCTCCTCAGA VIC: CATCCTGCTGGACCC 691 HR308691 275518708 R: AGACAGAGAAAGTCGGGACACT FAM: CATCCTGTTGGACCC Ots 103122-180 T/C F: CAAACGCGCACTCACACA VIC: CATCAACACAATCTGC 424 HR308692 275518709 R: TCACAATGGTACGATTTTACGACTCAA FAM: CATCAACACGATCTGC Ots 104048-194 T/C F: CAGCTGCTGCAGTCAATGAG VIC: CTGCCACCACCACCAC 383 HR308693 275518710 R: GCTCCTTACCAGTGTTGTCAGT FAM: TGCCACCGCCACCAC Ots 104063-132 C/T F: GCGTTACTGGTGTTATAAACGTTAGC VIC: CTTTCGTCCTTAGCACATAG 874 HR308694 275518711 R: GTTTATTTAATTATGAAGGACGATGTTGAAGTCA FAM: CTTTCGTCCTTAACACATAG Ots 104216-70 G/T F: AGTAGGAGTCGCAGCTATGGAA VIC: TCTGCCCCGGCTCT 546 HR308695 275518712 R: CCTGTGGTCGGAATGATGGT FAM: TCTGCCACGGCTCT Ots 104415-88 C/T F: CCTGAGCATCCCAGTTGAACT VIC: TCCTGAAAAACGACATCC 434 HR308696 275518713 R: TGTTTTCAATACACTGCAATTTAGTTTTGGT FAM: CTGAAAAACAACATCC Ots 104569-86 T/G F: CCTGCATGTTGTTCACGTTGTC VIC: TGGTCGCAGATGCC 582 HR308697 275518714 R: CGGCCGGAGGGATCAC FAM: TGGTCGCCGATGCC

25 Ots 105105-613 C/G F: AGTACAAGTGCAGAGAATGACATCATG VIC: CCGAGCTTGAGTTAGGA 801 HR308698 275518715 R: GGTGTTTTATTTTCCCATATATCTTTTAACTTTAAGCT FAM: CCGAGCTTGACTTAGGA Ots 105132-200 G/T F: CGATGTACTGAGGGCAGTGT VIC: CAAGAGTGGCATAAAA 458 HR308699 275518716 R: GAGTGGAGTTCCTTAATAATCATTGACCTT FAM: CAAGAGTGGAATAAAA Ots 105385-421 A/G F: GACTGTCTTGGAACCGTTGCTA VIC: CCTCCTGGGTATATCG 676 HR308700 275518717 R: TCCCGGAACACACCAATGTC FAM: CTCCTGGGCATATCG Ots 105401-325 T/G F: GAACTGAGCGGCTGCTG VIC: CAAGATGAGACAGTTACAG 500 HR308701 275518718 R: CGCCTCCTGGTGTCTATCCT FAM: CAAGATGAGACCGTTACAG Ots 105407-117 T/A F: TGTGTACATCCGCGTAAATATTGAAGATAA VIC: CAGGTTAGGAATGGTTG 476 HR308702 275518719 R: CTGTGAGCTGCTGCAAACC FAM: CAGGTTAGGATTGGTTG Ots 105897-124 T/C F: CCTCAGTGTTATTTGTATATGATCATTTTGAAACATTT VIC: AACCAATAATATGAAACTGTG 387 HR308703 275518720 R: AGCCCAATGCATCTACTGAATTCAT FAM: CCAATAATATGGAACTGTG Ots 106172-425 C/T F: GCAGTCAGTGCGTTGATACG VIC: CTGATAACTACTTGGCGTGTGT 466 HR308704 275518721 R: GGTGTAGACGTGAACAATGAGGATA FAM: TGATAACTACTTGGCATGTGT Ots 106313-729 G/A F: TTGTTCAATGGGCATTAATGCATGTT VIC: AAGAGTCCAGCGTTACTT 794 HR308705 275518722 R: TGCTTATGTGCAGATACTTGAGACAAA FAM: AAGAGTCCAGTGTTACTT Ots 106419b-618 T/G F: CAAGGGCACATTGGCAGATTTT VIC: CAATGATTAATGATTAATCCTTC 806 HR308706 275518723 R: ACCGGACCAAAGCACACA FAM: TGATTAATGATTCATCCTTC Ots 106499-70 C/G F: ACTCTATCATCGGCAGGACCAT VIC: CTCATTTTTCAGAATTGTATTC 516 HR308707 275518724 R: ACCGTAAGTGTGGTTGTGTTCATTA FAM: CTCATTTTTCAGAATTCTATTC Ots 106747-239 C/A F: ATCGAGGATGCCTCAAAGACATC VIC: CCCGCGGTGAGTAT 820 HR308708 275518725 R: GTTAGACCCACCACCAGTCATC FAM: CCCGCTGTGAGTAT Ots 107074-284 A/T F: CCCACTTCCAGAGCCTGAA VIC: ACCGTAGCTGCACCTG 399 HR308709 275518726 R: TTTTCCATGGCTGTGTGTACTGT FAM: CGTAGCAGCACCTG Continued on next page Table 1.2 – continued from previous page

Assay name Targets Primers (50-30) Probes (50-30) bp dbGSS dbSNP Ots 107220-70 C/G F: GGGAACGCACAACATACTACAGTA VIC: AGGTCATGAACTATAGATTCCATGTA 409 HR308710 275518727 R: AGAGATGAAACACTTCTATATCTATGGTGTGT FAM: TCATGAACTATAGATTCCATCTA Ots 107285-93 T/A F: GCCCTTGTGACAATGCACTGTTATA VIC: AAGTAACGTATCAAATGGC 726 HR308711 275518728 R: AACATACACCAATACTTAGGTCTAGACAGT FAM: AAAGTAACGTATCATATGGC Ots 107607-315 C/A F: GTGATGAGAGGTTTCCGGAAAATCT VIC: ATGGGAGACAGATAACT 516 HR308712 275518729 R: GTGTTCTGGATTCCATTGTGCAAA FAM: ATGGGAGACATATAACT Ots 107806-821 T/A F: CTCCCTTGCTTTTGGTCATTGG VIC: CAAAGAAAATCAAAATTT 997 HR308713 275518730 R: TGCAGTGCTGAATTAGAGATTAATTTTTGTG FAM: CAAAGAAAATCTAAATTT Ots 108007-208 A/T F: CAGGCTTGTGTTAAGTAGGGAGAAA VIC: CAGTTTCACTTAATTTTAAAATG 434 HR308714 275518731 R: CATTGGACAAGACCGGGTAGTC FAM: TTTCACTTAATTTAAAAATG Ots 108390-329 G/C F: GAGGTTTGTTACTGTCACCCATAGA VIC: CTACTTATGTAGCATTTTAA 1566 HR308715 275518732 R: CCTGCTGTAGCAAACTGTCTCAAA FAM: CTACTTATGTAGGATTTTAA Ots 108735-302 C/T F: CCTTTTTCTTATTAGTTTTACTTCCCCAGAGA VIC: AAACAAACAACGCCTCATG 324 HR308716 275518733 R: CAATTCCATTCTTGATTCTGTTTAACGGT FAM: AACAAACAACACCTCATG Ots 108820-336 G/A F: TGAAATAAATTGTTCTGTTGATATGTGAATTTTGGA VIC: ATTGCCCATCTCAGAATA 396 HR308717 275518734 R: CAACGACACACCAACAACGT FAM: AATTGCCCATCTTAGAATA Ots 109243-285 C/A F: CGCTGCTACACTGTTTACTGTGATA VIC: TGGCAAGATGACCTTT 514 HR308718 275518735 R: GCACCTCTTAAATTGTAAGTAAAATGTTAACGA FAM: AGTGGCAAGATTACCTTT Ots 109525-816 C/T F: GCCAGATAGTAGCGTACATCATGAG VIC: CATGAGGCGTTCGGC 1061 HR308719 275518736 R: CTCCCCATGTCCCTGAGTCT FAM: ATGAGGCATTCGGC

26 Ots 109693-392 T/G F: TCTCCCTCATTCCCATGTCATATCA VIC: TCCGTTAGTTCATCCTGG 473 HR308720 275518737 R: GGGAACGTATCAGGTGAGTGT FAM: TCCGTTAGTTCCTCCTGG Ots 110064-383 C/T F: AACAAAGAATGTTAAACACCAAACAGGAA VIC: CTACGTAATGAACGTTAGCT 801 HR308721 275518738 R: GTGCAAGGGACCTAGCTAATCC FAM: ACGTAATGAACATTAGCT Ots 110201-363 A/T F: GTTTGGCTATTGAAATTATACATTAAAACATGTAGCT VIC: TGGATGCCAGTTTTAAAA 558 HR308722 275518739 R: CCATGGCATCCTGTAAAGAACAACA FAM: TGGATGCCAGTTTAAAAA Ots 110381-164 A/G F: CTCTTGTTTGCTATGGGAGATGTAGT VIC: ATTTGCGTCTTCTCCC 661 HR308723 275518740 R: CCGTATCCTAAACCCTTCACTGTT FAM: TTGCGTCCTCTCCC Ots 110495-380 G/C F: GCCTAGGTATGTACGAAACTTCACA VIC: ATGGCCCCTGTCTATG 825 HR308724 275518741 R: AGGCTTTTTCAGATGGTCGTATGA FAM: ATGGCCCCTGTGTATG Ots 110551-64 C/A F: GAGTGGTCAAGGTTTCAGTTTCTG VIC: ACGCTCGGAACATT 685 HR308725 275518742 R: GAAATGGACAGACACAAGGTCAAAC FAM: ACGCTCTGAACATT Ots 110689-218 T/G F: GTATAAACTAGAGTCCAGTGTTATGTTAATGTCTT VIC: CACCAATCAATTAATTATT 397 HR308726 275518743 R: CATGGCAGACAACAGTAGAGAATATGA FAM: ACCAATCAATTCATTATT Ots 111084-96 A/G F: AAAAGTTAATACTGGGTACAACCTCTGAAAA VIC: TTGAACCATTCTACTATTGGT 710 HR308727 275518744 R: GGGACAGTAGTTGGGTCATTCAAAT FAM: AACCATTCTACCATTGGT Ots 111084b-619 C/A F: TTGTGGAATTACACCTTCAGAGTTCAAT VIC: CCATGGAAACGGACAAT 710 HR308727 275518745 R: GCCTGTTTGGCTTTCTTAAACTGAT FAM: TCCATGGAAACTGACAAT Ots 111312-435 C/T F: CCATGCGCCTTTGAGGAAATTAA VIC: ACTCATACCTAGAGGTCACAT 475 HR308728 275518746 R: TTCATGGTCTTTATCCCCCCTACA FAM: CTCATACCTAGAGATCACAT Ots 111666-408 C/T F: GAGAATCTGGGATTGGTACATCCAT VIC: ATAGTATCACTAGTTAAAAAT 664 HR308729 275518747 R: AAGCTCATGATACATGTATGAGTTATATTCTTCAAG FAM: ATAGTATCACTAATTAAAAAT Ots 111681-657 G/T F: CTGAGCTTTTTCAACTTACTTGTTGGA VIC: TAGCGCAAACCCCGAACC 702 HR308730 275518748 R: GGCGCAGCAGCAACTG FAM: CGCAAACACCGAACC Continued on next page Table 1.2 – continued from previous page

Assay name Targets Primers (50-30) Probes (50-30) bp dbGSS dbSNP Ots 112208-722 C/A F: CTGCATGAACGTTAACTCAAATAAAAGGT VIC: TGTGAGGGCGGTCTT 944 HR308731 275518749 R: AATGAGTTCTACTGACATTGTATACTAGAATAAGTATCA FAM: ATGTGAGGTCGGTCTT Ots 112301-43 T/C F: GCATGGCTGCCCTAGAACA VIC: CGTCGCATTCAGC 397 HR308732 275518750 R: TCAGAACATTTCCTTCAGCTTCGT FAM: CGTCGCGTTCAGC Ots 112419-131 A/T F: GTGGGTAATCGATGCCAAAGAGAT VIC: AAGCGACTTGATTATC 391 HR308733 275518751 R: TGGCAGTGTTTTCAACTAGCTTTG FAM: AGCGACATGATTATC Ots 112820-284 C/T F: CATAGATGTTTATATGAAAAACCTCCCACTGT VIC: ACTCACACTCGAGTGACT 394 HR308734 275518752 R: GCATCCAAAAAGACGTGTGTGTTT FAM: ACTCACACTCAAGTGACT Ots 112876-371 C/A F: GCCTACAGCAAATTCAGCTACACAT VIC: CATCACAACGATGTGTG 1118 HR308735 275518753 R: TGGACCTTCAATCATCACAGCTT FAM: CACATCACAACTATGTGTG Ots 113242-216 C/T F: GAGGCCTAATGTCTCTTGTGACT VIC: ATTACCAACGGAGAACC 364 HR308736 275518754 R: GACATCTTCAACAAGTGTTCATTCACC FAM: TTACCAACAGAGAACC Ots 113457-40 C/T F: CCCAAGTGGTGAGTGTCAGT VIC: ATATGGATTGGAGAATAG 555 HR308737 275518755 R: ACTACAACAGGTGTTGATAATAGAATCATTCTC FAM: CATATGGATTAGAGAATAG Ots 115987-325 T/G F: GGAGGTGTAGTGAAATGGGAAGAT VIC: ATGCATAAAAGGTAATTGTG 631 HR308738 275518756 R: GCATTCAGTGAACCAGTAGTGCTAT FAM: ATGCATAAAAGGTCATTGTG Ots 117043-255 C/A F: TCTCAATCTTGACACAAACTGGCT VIC: ACGTCAGAATGGATTCT 628 HR308739 275518757 R: TCGATCTGTTCTCGTGGTGTTTC FAM: AACGTCAGAATGTATTCT Ots 117138-545 T/G F: GGTGGTGGCAGTATTTGTTATCATG VIC: CAGTCAGACAGATACC 661 HR308740 275518758 R: GCAGTTACAGTCTGAGCTTGACAA FAM: CAGTCAGACCGATACC

27 Ots 117242-136 A/G F: GTGACAGGAGACAGAAAGAGACATT VIC: CAGCACATAACTTGACCTC 475 HR308741 275518759 R: TGGTCCTCCCTGTCTCTATCTACTA FAM: AGCACATAACCTGACCTC Ots 117259-271 T/G F: ACACCCACTTCAACCTCCATAAC VIC: CTCTCCTGATCACTCTGT 414 HR308742 275518760 R: GCCTCAGAGCTTAGCTTGGA FAM: CTCTCCTGATCCCTCTGT Ots 117370-471 T/G F: TGCAAACACAGAGGAAAGGGATTT VIC: ACGGAACAAATAAGACATTT 621 HR308743 275518761 R: GTTGGCTCCTTCAATTCAATTTGGA FAM: CGGAACAAATAAGCCATTT Ots 117432-409 A/G F: TCATCAAAACATGCCTCTTCTGTGT VIC: TTTAGACTTTGCTCTATAACAG 443 HR308744 275518762 R: TGTTGAACCTGTCACTCTGTCTTC FAM: ACTTTGCTCCATAACAG Ots 118175-479 C/T F: TGCGCGTCTCATTCAACCAT VIC: AGAATGAAGTGAAAAGAA 496 HR308745 275518763 R: ACCTTACGTCCTAGGTAGGAAACA FAM: AGAATGAAGTAAAAAGAA Ots 118205-61 T/C F: CCATACAGCCAGTCCAGGTG VIC: TAGTAGCCCCTACACCTC 485 HR308746 275518764 R: ACTGGACAGGGCTGGGT FAM: TAGCCCCTGCACCTC Ots 118938-325 C/T F: ATTTTCAAACAGGCATTTATCATTGGTGAA VIC: AGAGATGCAAAGTGGAGTT 606 HR308747 275518765 R: GGTCTGTCCCTCATTCTTTGCA FAM: AGAGATGCAAAATGGAGTT Ots 120950-417 T/A F: CAGACAGGTCACCATCACACT VIC: CTGGACCAGAACTCTGA 806 HR308748 275518766 R: TGGTGAAGCTGTAGGAGAAGGA FAM: CTGGACCAGATCTCTGA Ots 122414-56 C/T F: GCACCGTATCAACGAGCTCAT VIC: TGTATGACCTCTGACCTGT 423 HR308749 275518767 R: TGCATGGATTTCCTTTGTGTTGTTG FAM: TGTATGACCTCTAACCTGT Ots 123048-521 A/C F: CTCAACAGTGCACCTCCCTTAATT VIC: TCACATCCAACTCAGTACT 808 HR308750 275518768 R: CCAAACACACCCTTCCATAATCTCT FAM: CATCCAACGCAGTACT Ots 123205-61 G/A F: GCGCAAGAGGCGAAGATG VIC: CAAGTAGAATGCCTCCCCATA 329 HR308751 275518769 R: GGTCACAGCCATGGTGTTG FAM: CAAGTAGAATGCCTTCCCATA Ots 123921-111 A/G F: TCGCTAGGCAGAAATATAGGGTTCT VIC: TGCTAAATGGCATATATTAT 979 HR308752 275518770 R: GAGCATGGCGCTTGCA FAM: CTAAATGGCACATATTAT Continued on next page Table 1.2 – continued from previous page

Assay name Targets Primers (50-30) Probes (50-30) bp dbGSS dbSNP Ots 124774-477 T/C F: AGTTGTTCTTTTTATATTGTGTTTTTATTCCATTCCA VIC: CCACCGCCATCTGATA 710 HR308753 275518771 R: GCCAAATAAAAACAAAGCATGAACACA FAM: CACCGCCGTCTGATA Ots 126619-400 T/C F: GGATGGTTGTCATTTCTCTGCAAA VIC: AGAAAGTTCTAGAAATAATT 786 HR308754 275518772 R: CCGGGATACAATAATAATATTTGGTTAAGAGTTTTTT FAM: AAAGTTCTAGGAATAATT Ots 127236-62 T/A F: TGGAGAACTTGCACTGAATGTGAAA VIC: TCTCTTATCTGAGTTCTGC 668 HR308755 275518773 R: GCTGTTGGACCTTGACTTTAACAAATT FAM: CTCTTATCTGTGTTCTGC Ots 127760-569 C/T F: CTGCTGGCGCAGACATG VIC: CCGGTTTACCGATTTG 688 HR308756 275518774 R: CGTTATAGAGGATAGTTTGGAGGAAGGA FAM: CGGTTTACCAATTTG Ots 128302-57 C/T F: GGTTGCAGGGCAGAACTGT VIC: CCTGCAATACGACCAAC 833 HR308757 275518775 R: ACCCATCCAATAACCCATTTTCCTT FAM: CTGCAATACAACCAAC Ots 128495b-45 A/T F: GGACGCTGACCAGTACATAGG VIC: AGTGTCAACCTTCTCTC 153 HR308758 275518776 R: TTTCTCCCCAAAGTATTTAAGCCTACAC FAM: TGTCAACCATCTCTC Ots 128693-461 C/T F: TCAATGTTCATCAATGCACTTCCTGTA VIC: CACTCAGCTGGTACCCA 886 HR308759 275518777 R: GCCTGCAGGAGAAGGTAGAGTTA FAM: ACTCAGCTGATACCCA Ots 128757-61 A/G F: CGTGTCCGGCTTCTTTTATTTCATT VIC: TTGTGCATTTTCCCC 377 HR308760 275518778 R: GATGGGTATGTTAATCATATTACCAGCGTAA FAM: TGTGCATTTCCCCC Ots 129144-472 C/A F: CTGTTAGTGCAGAAGACGTAGCT VIC: TGGGTCTCGAGCCTGTA 635 HR308761 275518779 R: GCAGAGCTATTGAGCCAAGTTACAA FAM: TGGGTCTCGATCCTGTA Ots 129170-683 C/A F: AACCCTATGGGAACTCGTAGAACT VIC: ATTAGAACTCGTAGAACTAT 795 HR308762 275518780 R: GCTAGGAGTTCTCAAAAGGGTTCT FAM: ATATTAGAACTCGTATAACTAT

28 Ots 129303b-54 C/T F: ACCTGGAGAGAAAGTTCAAGAGAGA VIC: CCCTGGTGACCTCT 711 HR308763 275518781 R: GAGCTAGTAGAGGAAACAAAATAAACATTTCAAT FAM: CCCTGGTAACCTCT Ots 129458-451 T/C F: TGGGACCCACATAAAGCAACTG VIC: CATCTGGCAATGCCTT 551 HR308764 275518782 R: GACATAAGACCCATTTAGCCCCTTTT FAM: CATCTGGCAGTGCCTT Ots 129870-55 A/T F: GCATGTAACACATTATTTGGCATATGTACT VIC: ATGCATTCACCTGTATTAT 958 HR308765 275518783 R: CAGTACACTGGAGATTTGCAATGTT FAM: TGCATTCACCAGTATTAT Ots 130720-99 A/G F: CGGTCATTGTAAATGTCAACGGTTT VIC: CCTGTCTCATTCCC 542 HR308766 275518784 R: TGCTTGCATGTTCTTGGTGTAGTAA FAM: CTGTCCCATTCCC Ots 131460-584 T/C F: CCTATTTTTGATAGGTCATAGTGAATGGGATAG VIC: CTATCAAAGCAATACATTG 1283 HR308767 275518785 R: CTGTACTCCTCCATTCCTTTTCACT FAM: CTATCAAAGCAGTACATTG Ots 131802-393 T/C F: TGATTGTCTCATGGCCAATTGTCA VIC: TGTTCGAGAATGAAGATGAGTAA 489 HR308768 275518786 R: TGTAAATTCCACTTGGCAATCTTTGG FAM: TCGAGAATGAAGGTGAGTAA Ots 131906-141 A/T F: GGCTCGAACCACCCAGTTTA VIC: CACGGTTTACACTCCTATTA 408 HR308769 275518787 R: TGCCCAACTGGTTTGCAATC FAM: ACGGTTTACACTCCAATTA Ots AldB1-122 C/T F: GCCATGGAGGACTGGATGA VIC: ACCCACTTCGCCAACA 469 HR308770 275518788 R: GCCACCACTACTTGCTGAGAAAATA FAM: ACCCACTTCACCAACA Ots AldoB4-183 T/A F: TTTGTGCGTAAAGTCAGGTAGTGT VIC: CTGTGTGTCTAAGACAAT 296 HR308771 275518789 R: GTGCATGCCATGAGAACTTTGTTT FAM: CTGTGTGTCTATGACAAT Ots CathD-141 T/C F: CACTTGTTCTGCACACTACTTGTC VIC: TGGGAAGCAATCAA 484 HR308772 275518790 R: CACACATGGATTTTGCCTGTCTAAA FAM: AATTGGGAAGCAGTCAA Ots CRB-211 A/C F: CAACGCGGGAATGGCTTTTAA VIC: CTACCGTACTGAACTC 1041 HR308774 275518792 R: GCCAGAGTCGCCAAAATAGTAGAAT FAM: CCGTACGGAACTC Ots EndoRB1- G/A F: CCTTTGGGTCTGCTTGAGGTT VIC: TCCTTCTCACGCTTCT 1038 HR308775 275518793 486 R: GGAGCCAAATCCTAATGCTGAAGTA FAM: CTCCTTCTCATGCTTCT Continued on next page Table 1.2 – continued from previous page

Assay name Targets Primers (50-30) Probes (50-30) bp dbGSS dbSNP Ots Hsp90a G/C F: ACAGTATACCGGCTGCCTATTCATA VIC: ATTTGACTTGTCTTTTTG 373 HR308776 275518794 R: GTCGTTTTTCATAGAAAATAGCTCACAGTT FAM: TTTGACTTGTGTTTTTG Ots Myc-366 T/C F: CCTTAGCTGCTCTTTGAAGTTGACT VIC: TCTCTGCTCATCTGTC 409 HR308777 275518795 R: GGCTATAGAGTGTATTTACAGCATGCA FAM: CTCTGCTCGTCTGTC Ots ALDBINT1- T/C F: CGCTGGGCATGGATGAGT VIC: CTACTGTTGTATTTTCTC 474 HR308778 275518796 SNP1 R: GGCCAACACTGCTACTTCCT FAM: CTGTTGTGTTTTCTC Ots DESMIN19- C/A F: GGTCTGTCTGTCTGTCTATCTGTCA VIC: CCAGTCATGGGTCATT 439 HR308779 275518797 SNP1 R: TGTGTGTCTTTGTTCATTCCTACCA FAM: TCCAGTCATTGGTCATT Ots NAML12- A/G F: TGCCACCTCAGTTTTAGTGTTATATCC VIC: AAACCATTTTCATTCTTTTG 548 HR308780 275518798 SNP1 R: AGCGCCAACCTGTCACT FAM: CCATTTTCACTCTTTTG Ots NAML12- C/A F: GGCGGTTAGGTAGGATATGATTCC VIC: TCCATAAGCGGGAAAA 548 HR308780 275518799 SNP2 R: TCACGTAGCCTACCACAGATAAGT FAM: TTTCCATAAGCGTGAAAA Ots BMP2- C/T F: ACTGCCACAGACACGAACTC VIC: CCCACTTCGCTGAAGT 638 HR308781 275518800 SNP1 R: GCCACTATCCACTCGTTCCA FAM: CCCACTTCACTGAAGT Ots MTA- C/T F: GCCGAAAAATAAGCGATTAGTGATGA VIC: AATTGCCTCATTGGGTG 220 HR308782 275518801 SNP1 R: GCCCCATGGTAAACCTAATTAACCT FAM: AATTGCCTCATTAGGTG Ots TF1- G/T F: CGGACAAAGAGCTACAGAAATGC VIC: CCGCCACCTTGGCT 755 HR308783 275518802 SNP1 R: CGTCCCTCTTCACGCATGA FAM: CGCCACATTGGCT 29 All TaqMan assays were then used to genotype 337 ﬁsh from ﬁve major salmon stocks representing the Sacramento (Central Valley), Klamath and Columbia Rivers.

A summary of the population genetic variability of the 117 validated assays can be found in Table 1.3. Mean MAF for all of the variable loci ranged from 0.194 in the

Klamath/Trinity basin to 0.236 in the Feather River Spring-run stock. MAFs for individual loci in these ﬁve populations ranged from to 0.005 to 0.500. The proportion of polymorphic loci was nearly 90% in all populations and ranged from 90.6% in the lower Columbia stock to 86.3% in the Butte Creek Spring-run population. Thirty of the

117 loci were monomorphic in at least one population and four loci (Ots 102195-157,

Ots 107220-70, Ots 117138-545, and Ots 123205-61) were not variable in any of the samples from the ﬁve populations/stocks. They are reported here, however, because either the ascertainment sequence data or additional genotype data (not shown) indicate these markers are variable in the species and may be useful in other parts of the range.

Expected (unbiased) heterozygosity (HE) for each variable locus ranged from 0.01 to

0.51 (mean = 0.33), while observed heterozygosity (HO) ranged from 0.01 to 0.68 (mean

= 0.33). Mean HE was similar in all populations, ranging from 0.27 (Klamath) to 0.31

(Feather River and lower Columbia), and mean HO followed the same pattern, ranging from 0.26 (Butte Creek) to 0.31 (Feather River). Overall FST for the individual loci and

all ﬁve populations ranged from 0 to 0.592 and averaged 0.107 for all loci, indicating

substantial diﬀerentiation in allele frequencies. Almost all loci were in Hardy Wein-

berg equilibrium in all populations; only two loci (Ots 127760-569 and Ots 109243-285) deviated from equilibrium in one and two populations, respectively (Table 1.3).

30 Table 1.3: Summary statistics for 117 SNP loci in five Chinook salmon populations. N is the number of individuals genotyped. HE is expected (unbiased) heterozygosity and HO is observed heterozygosity. FST is over all five populations. AF is the observed frequency of the minor allele from the Feather River stock in each population. Asterisks (*) indicate significant (p<0.001) deviations from Hardy-Weinberg equilibrium.

Feather River Butte Creek Central Valley Klam/Trinity L. Columbia (Spring) (Spring) (Fall) (Fall) (Spring) N=94 N=54 N=94 N=48 N=47

Assay AF HE HO AF HE HO AF HE HO AF HE HO AF HE HO FST Ots 94857-232 0.466 0.50 0.42 0.567 0.50 0.56 0.582 0.49 0.44 0.222 0.35 0.36 0.330 0.45 0.57 0.076 Ots 94903-99 0.158 0.27 0.25 0.078 0.15 0.12 0.137 0.24 0.25 0.010 0.02 0.02 0.255 0.38 0.47 0.048 Ots 95442b-204 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.030 0.06 0.07 0.150 0.26 0.21 0.124 Ots 96222-525 0.375 0.47 0.51 0.549 0.50 0.39 0.344 0.45 0.47 0.848 0.26 0.30 0.522 0.50 0.57 0.135 Ots 96500-180 0.194 0.32 0.32 0.149 0.26 0.21 0.096 0.17 0.15 0.622 0.48 0.32 0.315 0.44 0.33 0.183

31 Ots 96899-357 0.022 0.04 0.04 0.000 0.00 0.00 0.016 0.03 0.03 0.000 0.00 0.00 0.000 0.00 0.00 0.004 Ots 97077-179 0.289 0.41 0.44 0.104 0.19 0.08 0.190 0.31 0.34 0.239 0.37 0.35 0.178 0.30 0.22 0.021 Ots 97660-56 0.049 0.09 0.10 0.010 0.02 0.02 0.027 0.05 0.05 0.010 0.02 0.02 0.011 0.02 0.02 0.006 Ots 98409-850 0.054 0.10 0.11 0.031 0.06 0.06 0.075 0.14 0.13 0.000 0.00 0.00 0.000 0.00 0.00 0.021 Ots 98683-796 0.112 0.20 0.16 0.120 0.21 0.24 0.081 0.15 0.16 0.052 0.10 0.10 0.053 0.10 0.11 0.003 Ots 99550-204 0.350 0.46 0.48 0.202 0.33 0.37 0.287 0.41 0.38 0.245 0.37 0.40 0.163 0.28 0.28 0.020 Ots 100884-287 0.081 0.15 0.14 0.051 0.10 0.10 0.114 0.20 0.18 0.318 0.44 0.36 0.141 0.25 0.24 0.067 Ots 101119-381 0.429 0.49 0.53 0.573 0.49 0.40 0.544 0.50 0.44 0.032 0.06 0.06 0.064 0.12 0.13 0.232 Ots 101554-407 0.174 0.29 0.26 0.100 0.18 0.20 0.183 0.30 0.30 0.010 0.02 0.02 0.223 0.35 0.36 0.037 Ots 101704-143 0.367 0.47 0.46 0.696 0.43 0.39 0.456 0.50 0.43 0.659 0.45 0.41 0.426 0.49 0.43 0.069 Ots 101770-82 0.255 0.38 0.36 0.010 0.02 0.02 0.242 0.37 0.31 0.010 0.02 0.02 0.000 0.00 0.00 0.140 Ots 102195-157 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 Ots 102213-210 0.228 0.35 0.39 0.138 0.24 0.28 0.196 0.32 0.37 0.240 0.37 0.40 0.021 0.04 0.04 0.038 Ots 102414-395 0.440 0.50 0.48 0.604 0.48 0.46 0.382 0.47 0.46 0.628 0.47 0.66 0.479 0.50 0.45 0.037 Ots 102420-494 0.333 0.45 0.51 0.220 0.35 0.36 0.277 0.40 0.45 0.010 0.02 0.02 0.422 0.49 0.49 0.088 Ots 102457-132 0.170 0.28 0.32 0.112 0.20 0.18 0.242 0.37 0.42 0.708 0.42 0.42 0.522 0.50 0.30 0.234 Ots 102801-308 0.261 0.39 0.41 0.344 0.46 0.44 0.258 0.39 0.34 0.309 0.43 0.57 0.163 0.28 0.28 0.010 Continued on next page Table 1.3 – continued from previous page

Feather River Butte Creek Central Valley Klam/Trinity L. Columbia (Spring) (Spring) (Fall) (Fall) (Spring) N=94 N=54 N=94 N=48 N=47

Assay AF HE HO AF HE HO AF HE HO AF HE HO AF HE HO FST Ots 102867-609 0.462 0.50 0.48 0.373 0.47 0.51 0.376 0.47 0.49 0.948 0.10 0.10 0.678 0.44 0.47 0.193 Ots 103041-52 0.494 0.50 0.58 0.337 0.45 0.31 0.350 0.46 0.38 0.351 0.46 0.53 0.435 0.50 0.52 0.015 Ots 103122-180 0.005 0.01 0.01 0.020 0.04 0.04 0.000 0.00 0.00 0.174 0.29 0.30 0.522 0.50 0.42 0.404 Ots 104048-194 0.191 0.31 0.36 0.225 0.35 0.45 0.242 0.37 0.48 0.083 0.15 0.17 0.109 0.20 0.22 0.024 Ots 104063-132 0.054 0.10 0.11 0.179 0.30 0.26 0.038 0.07 0.05 0.436 0.50 0.57 0.128 0.23 0.21 0.188 Ots 104216-70 0.000 0.00 0.00 0.000 0.00 0.00 0.011 0.02 0.02 0.083 0.15 0.17 0.032 0.06 0.06 0.046 Ots 104415-88 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.010 0.02 0.02 0.096 0.18 0.19 0.086 Ots 104569-86 0.500 0.50 0.45 0.656 0.46 0.27 0.419 0.49 0.47 0.478 0.50 0.57 0.391 0.48 0.39 0.027 Ots 105105-613 0.495 0.50 0.53 0.522 0.50 0.48 0.690 0.43 0.39 0.128 0.23 0.26 0.478 0.50 0.51 0.145

32 Ots 105132-200 0.209 0.33 0.29 0.070 0.13 0.06 0.210 0.33 0.33 0.083 0.15 0.13 0.022 0.04 0.04 0.052 Ots 105385-421 0.054 0.10 0.11 0.060 0.11 0.12 0.038 0.07 0.08 0.000 0.00 0.00 0.245 0.37 0.45 0.099 Ots 105401-325 0.239 0.37 0.37 0.275 0.40 0.35 0.324 0.44 0.43 0.143 0.25 0.19 0.156 0.27 0.27 0.022 Ots 105407-117 0.348 0.46 0.41 0.510 0.50 0.58 0.277 0.40 0.47 0.073 0.14 0.15 0.160 0.27 0.28 0.102 Ots 105897-124 0.122 0.22 0.20 0.102 0.19 0.20 0.090 0.16 0.16 0.198 0.32 0.23 0.011 0.02 0.02 0.028 Ots 106172-425 0.000 0.00 0.00 0.083 0.15 0.17 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.084 Ots 106313-729 0.148 0.25 0.25 0.051 0.10 0.06 0.133 0.23 0.20 0.415 0.49 0.53 0.189 0.31 0.33 0.093 Ots 106419b-618 0.163 0.27 0.28 0.225 0.35 0.29 0.092 0.17 0.16 0.021 0.04 0.04 0.489 0.51 0.41 0.160 Ots 106499-70 0.304 0.43 0.37 0.163 0.28 0.24 0.237 0.36 0.34 0.587 0.49 0.52 0.457 0.50 0.49 0.098 Ots 106747-239 0.339 0.45 0.47 0.310 0.43 0.38 0.393 0.48 0.40 0.239 0.37 0.39 0.553 0.50 0.47 0.037 Ots 107074-284 0.429 0.49 0.49 0.333 0.45 0.38 0.378 0.47 0.44 0.677 0.44 0.52 0.544 0.50 0.35 0.056 Ots 107220-70 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 Ots 107285-93 0.233 0.36 0.42 0.480 0.50 0.60 0.194 0.31 0.34 0.138 0.24 0.15 0.156 0.27 0.18 0.077 Ots 107607-315 0.255 0.38 0.40 0.177 0.29 0.24 0.236 0.36 0.43 0.117 0.21 0.23 0.261 0.39 0.30 0.011 Ots 107806-821 0.208 0.33 0.35 0.323 0.44 0.44 0.294 0.42 0.41 0.128 0.23 0.17 0.330 0.45 0.39 0.025 Ots 108007-208 0.317 0.44 0.43 0.202 0.33 0.37 0.275 0.40 0.46 0.652 0.46 0.52 0.422 0.49 0.31 0.098 Continued on next page Table 1.3 – continued from previous page

Feather River Butte Creek Central Valley Klam/Trinity L. Columbia (Spring) (Spring) (Fall) (Fall) (Spring) N=94 N=54 N=94 N=48 N=47

Assay AF HE HO AF HE HO AF HE HO AF HE HO AF HE HO FST Ots 108390-329 0.444 0.50 0.48 0.448 0.50 0.44 0.353 0.46 0.47 0.630 0.47 0.57 0.234 0.36 0.38 0.060 Ots 108735-302 0.312 0.43 0.48 0.132 0.23 0.16 0.371 0.47 0.46 0.500 0.51 0.50 0.326 0.44 0.51 0.050 Ots 108820-336 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.362 0.47 0.51 0.362 0.47 0.43 0.337 Ots 109243-285 0.489 0.50* 0.35 0.511 0.51 0.43 0.550 0.50 0.37 0.630 0.47 0.35 0.500 0.51* 0.24 0.001 Ots 109525-816 0.000 0.00 0.00 0.000 0.00 0.00 0.005 0.01 0.01 0.031 0.06 0.06 0.032 0.06 0.06 0.015 Ots 109693-392 0.081 0.15 0.16 0.020 0.04 0.04 0.038 0.07 0.05 0.532 0.50 0.51 0.383 0.48 0.43 0.302 Ots 110064-383 0.500 0.50 0.48 0.548 0.50 0.37 0.429 0.49 0.47 0.609 0.48 0.43 0.554 0.50 0.54 0.011 Ots 110201-363 0.380 0.47 0.54 0.217 0.34 0.35 0.451 0.50 0.48 0.281 0.41 0.48 0.261 0.39 0.35 0.035 Ots 110381-164 0.194 0.32 0.34 0.194 0.32 0.35 0.278 0.40 0.38 0.348 0.46 0.57 0.294 0.42 0.24 0.014

33 Ots 110495-380 0.411 0.49 0.44 0.271 0.40 0.38 0.450 0.50 0.57 0.106 0.19 0.17 0.348 0.46 0.39 0.067 Ots 110551-64 0.244 0.37 0.36 0.200 0.32 0.32 0.178 0.29 0.31 0.000 0.00 0.00 0.202 0.33 0.32 0.045 Ots 110689-218 0.301 0.42 0.40 0.120 0.21 0.15 0.368 0.47 0.55 0.330 0.45 0.45 0.128 0.23 0.26 0.055 Ots 111084-96 0.484 0.50 0.44 0.568 0.50 0.36 0.467 0.50 0.47 0.844 0.27 0.22 0.815 0.30 0.33 0.118 Ots 111084b-619 0.000 0.00 0.00 0.019 0.04 0.04 0.000 0.00 0.00 0.067 0.13 0.13 0.261 0.39 0.34 0.196 Ots 111312-435 0.000 0.00 0.00 0.000 0.00 0.00 0.006 0.01 0.01 0.556 0.50 0.28 0.171 0.29 0.13 0.456 Ots 111666-408 0.378 0.47 0.52 0.455 0.50 0.50 0.357 0.46 0.43 0.958 0.08 0.08 0.144 0.25 0.24 0.267 Ots 111681-657 0.435 0.49 0.52 0.327 0.44 0.49 0.602 0.48 0.52 0.878 0.22 0.24 0.576 0.49 0.41 0.130 Ots 112208-722 0.494 0.50 0.41 0.413 0.49 0.48 0.432 0.49 0.48 0.915 0.16 0.17 0.389 0.48 0.42 0.140 Ots 112301-43 0.121 0.21 0.24 0.080 0.15 0.16 0.088 0.16 0.13 0.521 0.50 0.57 0.370 0.47 0.48 0.198 Ots 112419-131 0.344 0.45 0.42 0.673 0.44 0.35 0.452 0.50 0.39 0.532 0.50 0.55 0.394 0.48 0.45 0.054 Ots 112820-284 0.462 0.50 0.65 0.245 0.37 0.41 0.456 0.50 0.47 0.870 0.23 0.22 0.733 0.40 0.36 0.179 Ots 112876-371 0.478 0.50 0.52 0.310 0.43 0.42 0.360 0.46 0.42 0.222 0.35 0.31 0.457 0.50 0.39 0.034 Ots 113242-216 0.396 0.48 0.44 0.630 0.47 0.46 0.467 0.50 0.56 0.444 0.50 0.67 0.359 0.47 0.54 0.030 Ots 113457-40 0.233 0.36 0.38 0.226 0.35 0.22 0.299 0.42 0.32 0.032 0.06 0.06 0.217 0.34 0.35 0.043 Ots 115987-325 0.000 0.00 0.00 0.000 0.00 0.00 0.016 0.03 0.03 0.723 0.40 0.43 0.489 0.51 0.49 0.592 Continued on next page Table 1.3 – continued from previous page

Feather River Butte Creek Central Valley Klam/Trinity L. Columbia (Spring) (Spring) (Fall) (Fall) (Spring) N=94 N=54 N=94 N=48 N=47

Assay AF HE HO AF HE HO AF HE HO AF HE HO AF HE HO FST Ots 117043-255 0.460 0.50 0.53 0.423 0.49 0.46 0.452 0.50 0.54 0.798 0.33 0.36 0.833 0.28 0.24 0.132 Ots 117138-545 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 Ots 117242-136 0.447 0.50 0.45 0.143 0.25 0.20 0.456 0.50 0.54 0.229 0.36 0.46 0.207 0.33 0.28 0.089 Ots 117259-271 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.115 0.21 0.19 0.213 0.34 0.30 0.163 Ots 117370-471 0.169 0.28 0.31 0.040 0.08 0.08 0.275 0.40 0.33 0.128 0.23 0.26 0.085 0.16 0.13 0.057 Ots 117432-409 0.444 0.50 0.49 0.290 0.42 0.46 0.318 0.44 0.48 0.128 0.23 0.17 0.294 0.42 0.37 0.051 Ots 118175-479 0.170 0.28 0.32 0.245 0.37 0.29 0.270 0.40 0.36 0.766 0.36 0.43 0.239 0.37 0.43 0.211 Ots 118205-61 0.048 0.09 0.10 0.441 0.50 0.41 0.049 0.09 0.10 0.042 0.08 0.08 0.196 0.32 0.35 0.216 Ots 118938-325 0.183 0.30 0.21 0.163 0.28 0.33 0.196 0.32 0.35 0.063 0.12 0.13 0.652 0.46 0.57 0.206

34 Ots 120950-417 0.106 0.19 0.21 0.147 0.25 0.14 0.122 0.22 0.24 0.073 0.14 0.10 0.064 0.12 0.13 0.002 Ots 122414-56 0.425 0.49 0.48 0.387 0.48 0.51 0.226 0.35 0.34 0.611 0.48 0.60 0.630 0.47 0.61 0.107 Ots 123048-521 0.483 0.50 0.56 0.365 0.47 0.38 0.359 0.46 0.48 0.000 0.00 0.00 0.130 0.23 0.26 0.156 Ots 123205-61 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 0.00 0.00 0.000 Ots 123921-111 0.272 0.40 0.41 0.106 0.19 0.21 0.222 0.35 0.31 0.000 0.00 0.00 0.064 0.12 0.13 0.086 Ots 124774-477 0.467 0.50 0.54 0.309 0.43 0.36 0.489 0.50 0.53 0.271 0.40 0.42 0.337 0.45 0.41 0.034 Ots 126619-400 0.406 0.49 0.46 0.276 0.40 0.39 0.353 0.46 0.38 0.359 0.47 0.37 0.319 0.44 0.38 0.001 Ots 127236-62 0.205 0.33 0.30 0.150 0.26 0.26 0.247 0.37 0.43 0.958 0.08 0.08 0.330 0.45 0.40 0.362 Ots 127760-569 0.023 0.04* 0.04 0.011 0.02 0.02 0.000 0.00 0.00 0.000 0.00 0.00 0.075 0.14 0.11 0.034 Ots 128302-57 0.286 0.41 0.40 0.150 0.26 0.26 0.165 0.28 0.24 0.447 0.50 0.47 0.138 0.24 0.15 0.070 Ots 128495b-45 0.352 0.46 0.46 0.457 0.50 0.45 0.397 0.48 0.43 0.261 0.39 0.34 0.467 0.50 0.40 0.014 Ots 128693-461 0.304 0.43 0.39 0.500 0.51 0.36 0.500 0.50 0.49 0.457 0.50 0.53 0.319 0.44 0.34 0.034 Ots 128757-61 0.473 0.50 0.62 0.451 0.50 0.51 0.450 0.50 0.50 0.115 0.21 0.23 0.394 0.48 0.53 0.070 Ots 129144-472 0.346 0.46 0.47 0.451 0.50 0.59 0.332 0.45 0.40 0.000 0.00 0.00 0.117 0.21 0.23 0.129 Ots 129170-683 0.247 0.37 0.34 0.118 0.21 0.24 0.137 0.24 0.23 0.458 0.50 0.42 0.170 0.29 0.21 0.084 Ots 129303b-54 0.407 0.45 0.43 0.471 0.27 0.24 0.362 0.41 0.33 0.083 0.51 0.51 0.245 0.44 0.49 0.043 Continued on next page Table 1.3 – continued from previous page

Feather River Butte Creek Central Valley Klam/Trinity L. Columbia (Spring) (Spring) (Fall) (Fall) (Spring) N=94 N=54 N=94 N=48 N=47

Assay AF HE HO AF HE HO AF HE HO AF HE HO AF HE HO FST Ots 129458-451 0.495 0.50 0.56 0.635 0.47 0.40 0.451 0.50 0.51 0.625 0.47 0.48 0.380 0.48 0.50 0.031 Ots 129870-55 0.299 0.42 0.36 0.337 0.45 0.47 0.301 0.42 0.41 0.021 0.04 0.04 0.160 0.27 0.28 0.070 Ots 130720-99 0.093 0.17 0.19 0.000 0.00 0.00 0.098 0.18 0.17 0.333 0.45 0.49 0.402 0.49 0.28 0.166 Ots 131460-584 0.170 0.28 0.27 0.120 0.21 0.24 0.231 0.36 0.37 0.322 0.44 0.56 0.457 0.50 0.40 0.071 Ots 131802-393 0.214 0.34 0.34 0.240 0.37 0.32 0.225 0.35 0.38 0.021 0.04 0.04 0.075 0.14 0.15 0.052 Ots 131906-141 0.100 0.18 0.18 0.220 0.35 0.32 0.112 0.20 0.22 0.692 0.43 0.53 0.163 0.28 0.24 0.278 Ots AldB1-122 0.113 0.20 0.20 0.210 0.34 0.22 0.082 0.15 0.14 0.635 0.47 0.52 0.021 0.04 0.04 0.299 Ots AldoB4-183 0.006 0.01 0.01 0.020 0.04 0.04 0.000 0.00 0.00 0.489 0.51 0.68 0.000 0.00 0.00 0.475 Ots CathD-141 0.016 0.03 0.03 0.010 0.02 0.02 0.000 0.00 0.00 0.304 0.43 0.61 0.000 0.00 0.00 0.275

35 Ots CRB-211 0.005 0.01 0.01 0.000 0.00 0.00 0.005 0.01 0.01 0.096 0.18 0.15 0.000 0.00 0.00 0.072 Ots EndoRB1-486 0.437 0.50 0.67 0.659 0.45 0.41 0.278 0.40 0.36 0.010 0.02 0.02 0.267 0.40 0.44 0.194 Ots Hsp90a 0.016 0.03 0.03 0.000 0.00 0.00 0.067 0.13 0.13 0.188 0.31 0.29 0.337 0.45 0.28 0.172 Ots Myc-366 0.258 0.39 0.43 0.100 0.18 0.12 0.211 0.33 0.36 0.000 0.00 0.00 0.011 0.02 0.02 0.100 Ots ALDBINT1-SNP1 0.133 0.23 0.24 0.147 0.25 0.22 0.192 0.31 0.32 0.734 0.39 0.32 0.750 0.38 0.41 0.369 Ots DESMIN19-SNP1 0.218 0.34 0.37 0.170 0.29 0.17 0.207 0.33 0.28 0.174 0.29 0.30 0.149 0.26 0.21 -0.004 Ots NAML12-SNP1 0.197 0.32 0.28 0.250 0.38 0.42 0.177 0.29 0.27 0.053 0.10 0.11 0.386 0.48 0.36 0.058 Ots NAML12-SNP2 0.209 0.33 0.37 0.096 0.18 0.15 0.247 0.37 0.36 0.178 0.30 0.31 0.043 0.08 0.09 0.039 Ots BMP2-SNP1 0.319 0.44 0.40 0.357 0.46 0.43 0.204 0.33 0.28 0.234 0.36 0.30 0.054 0.10 0.11 0.054 Ots MTA-SNP1 0.239 0.37 0.35 0.271 0.40 0.38 0.390 0.48 0.49 0.042 0.08 0.08 0.266 0.39 0.40 0.071 Ots TF1-SNP1 0.111 0.20 0.22 0.214 0.34 0.27 0.132 0.23 0.20 0.367 0.47 0.60 0.436 0.50 0.49 0.106 Mean 0.236 0.31 0.31 0.203 0.28 0.26 0.224 0.30 0.30 0.194 0.27 0.28 0.233 0.31 0.29 0.107 Polymorphic Loci (%) 88.0 86.3 88.0 88.0 90.6 While full annotation of these gene fragments is beyond the scope of the present study, preliminary BLAST (Basic Local Alignment Search Tool, NCBI) results and annotation of the target SNP appear in Table 1.4. Note that we have included annotation not just for the loci described here (Reference 1), but also for an additional 24 loci

(References 2, 3, 4 and unpublished) that are part of the ﬁnal genotyping panel described in Chapter 2. To determine whether the target variation was in an intron or an exon, we aligned the genomic sequence from our Sanger sequencing eﬀort with the

EST sequence from which initial primers were designed. Over all 141 loci, 81 SNPs were found in exons while 48 SNPs were found in introns. For 12 loci, EST sequence was unavailable or gene annotation was insuﬃcient to determine intron/exon boundaries.

BLAST results revealed identity with a known gene or genomic fragment for 92 loci with E-values ranging from 1.2E-11 to zero (smaller numbers indicating higher similarity). The target SNP was found to be in the 5’ or 3’ untranslated region (UTR) of a gene for 25 loci, while an annotated translation (n.t.) was unavailable for 10 loci.

Introns were found in a variety of locations with respect to the described gene or gene fragment; the introns for 24 loci were found within the coding sequence (CDS) of a gene while other were found up or downstream of the annotated region. Two of the described

SNPs are within microsatellite (msat) repeats. Of the variation found in CDS exons, nine were synonymous substitutions while nine represented mutations that altered the resulting amino acid at that position (nonsynonymous); the status of the CDS SNP at

Ots AldoB4-183 could not be determined due to poor sequence homology.

36 Table 1.4: Preliminary BLAST results (BLAST hit and e-value) and annotation of the target SNP for the loci described here (Reference 1) and for an additional 24 loci (References 2, 3, 4 and unpublished) that are part of the ﬁnal genotyping panel described in Chapter 2. Also included is whether the variation is present in an intron or exon and its location with respect to the described gene, either in coding sequence (CDS) or untranslated regions (UTR). No translation (n.t.) was available for 10 loci. For CDS exons, a single amino acid is indicated for synonymous substitutions, while both amino acids are included for non-synonymous substitutions. Reference codes are as follows: 1. Clemento et al. 2011; 2. Smith et al. 2005a; 3. Campbell and Narum 2008; 4. Smith et al. 2005b.

Assay name Ref SNP Genic location BLAST # [E-value] Description Ots 94857-232 1 intron in MSAT repeat AY543888 [3.8E-18] Salmo salar clone Alu374 microsatellite sequence Ots 94903-99 1 exon Ots 95442b-204 1 intron CDS NM 001123604 [0] Salmo salar somatolactin (LOC100136491) Ots 96222-525 1 intron CDS AB326306 [2.02E-45] Solea senegalensis elongation factor 1 alpha isoform 42Sp50 Ots 96500-180 1 exon Ots 96899-357 1 intron CDS NM 001165332 [5.1E-56] Salmo salar vaccinia related kinase 3 (vrk3) Ots 97077-179 1 exon 3’ UTR BT045669 [3.04E-100] Salmo salar BTG Ots 97660-56 1 exon 5’ of gene EU025717 [7.07E-29] Salmo salar single-strand selective monofunctional uracil 37 Ots 98409-850 1 intron CDS BT072224 [1.01E-45] Salmo salar ﬁzzy-related protein homolog Ots 98683-796 1 intron CDS NM 001141554 [1.32E-83] Salmo salar chymotrypsin-like (ctrl) Ots 99550-204 1 exon n.t. (pseudogene) BT071884 [4.88E-69] Salmo salar collagen alpha-2VI chain precursor Ots 100884-287 1 exon 3’UTR NM 001140998 [7.34E-173] Salmo salar centrosomal protein 97 (cep97) Ots 101119-381 1 intron Ots 101554-407 1 exon 3’UTR BT058832 [8.48E-177] Salmo salar NMDA receptor-regulated protein 1 Ots 101704-143 1 exon n.t. (pseudogene) BT071847 [3.89E-97] Salmo salar histidyl-tRNA synthetase Ots 101770-82 1 exon CDS [Pro] BT078766 [5.18E-77] Esox lucius cartilage-associated protein precursor Ots 102195-157 1 exon CDS [Pro>Ser] NM 001160495 [1.99E-119] O. mykiss C type lectin receptor B Ots 102213-210 1 intron CDS BT073541 [5.62E-76] O. mykiss leukocyte cell-derived chemotaxin 2 precursor Ots 102414-395 1 intron Ots 102420-494 1 exon 3’UTR BT058671 [1.01E-123] Salmo salar cathepsin K precursor Ots 102457-132 1 exon 3’UTR (STS) NM 001124228 [2.41E-64] O. mykiss heat shock protein 70a (hsp70a) Ots 102801-308 1 intron Ots 102867-609 1 exon Ots 103041-52 1 exon Ots 103122-180 1 exon Ots 104048-194 1 exon Ots 104063-132 1 intron 5’ of gene NM 001160489 [0] O. mykiss mitochondrial complex I subunit NDUFB2 Ots 104216-70 1 exon Ots 104415-88 1 exon 3’UTR NM 001140438 [1.54E-134] Salmo salar asparagine-linked glycosylation 12 homolog Continued on next page Table 1.4 – continued from previous page

Assay name Ref SNP Genic location BLAST # [E-value] Description Ots 104569-86 1 exon Ots 105105-613 1 exon CDS [Thr>Ser] AF483538 [9.12E-90] O. mykiss VHSV-induced protein mRNA Ots 105132-200 1 exon 3’ of gene NM 001173959 [1.33E-139] Salmo salar member of RAS oncogene family (rap2c) Ots 105385-421 1 exon Ots 105401-325 1 exon n.t. ET306160 [9.74E-64] Salmo salar genomic clone S0262B04 Ots 105407-117 1 exon 5’UTR BT047755 [8.13E-133] Salmo salar 60S ribosomal protein L36a Ots 105897-124 1 exon 3’UTR NM 001173890 [2.35E-127] Salmo salar uridine 5-monophosphate synthase (pyr5) Ots 106172-425 1 exon Ots 106313-729 1 intron CDS NM 001141707 [9.98E-129] Salmo salar Wilms tumor 1 associated protein-like Ots 106419b-618 1 intron 5’ of gene AF281332 [5.35E-77] O. mykiss biotinidase fragment 1 mRNA Ots 106499-70 1 intron CDS NM 001124329 [1.2E-11] O. mykiss superoxide dismutase 1 (sod1) Ots 106747-239 1 exon CDS [Pro>Gln] NM 001140141 [8.43E-95] Salmo salar C10orf88 homolog Ots 107074-284 1 exon Ots 107220-70 1 exon Ots 107285-93 1 intron CDS NM 001146578 [1.05E-73] Salmo salar placental protein 25 Ots 107607-315 1 exon Ots 107806-821 1 intron CDS BT048833 [1.89E-82] Salmo salar YIPF4 38 Ots 108007-208 1 exon Ots 108390-329 1 intron 3’UTR BT045571 [3.92E-56] Salmo salar ADP-ribosylation factor-like protein 9 Ots 108735-302 1 exon Ots 108820-336 1 exon Ots 109243-285 1 exon 3’UTR NM 001140063 [2.15E-59] Salmo salar Kunitz-type protease inhibitor 2 (spit2) Ots 109525-816 1 exon CDS [Ala] NM 001124667 [1.46E-128] O. mykiss prostaglandin synthase 2b (ptgs2b) Ots 109693-392 1 exon 3’UTR NM 001139789 [0] Salmo salar nuclear transcription factor Y subunit gamma Ots 110064-383 1 intron CDS NM 001165121 [4.29E-73] O. mykiss lipopolysaccharide-induced tnf-alpha Ots 110201-363 1 intron Ots 110381-164 1 exon Ots 110495-380 1 intron Ots 110551-64 1 exon n.t. AL954310 [8.23E-35] Danio rerio DNA sequence in linkage group 8 Ots 110689-218 1 exon Ots 111084-96 1 exon Ots 111084b-619 1 intron Ots 111312-435 1 exon n.t. (pseudogene) BT072385 [0] Salmo salar DNA topoisomerase 1 Ots 111666-408 1 intron CDS NM 001140911 [2.1E-50] Salmo salar Tetraspanin-16 (tsn16) Ots 111681-657 1 exon CDS [Thr>Pro] NM 001124224 [0] O. mykiss NK2 homeobox 1b (nkx2.1b) Ots 112208-722 1 intron CDS DQ784539 [0] O. mykiss CW lactate dehydrogenase B gene Ots 112301-43 1 intron Ots 112419-131 1 exon 3’UTR NM 001139960 [2.38E-107] Salmo salar junction plakoglobin (plak) Continued on next page Table 1.4 – continued from previous page

Assay name Ref SNP Genic location BLAST # [E-value] Description Ots 112820-284 1 exon Ots 112876-371 1 intron CDS HE608241 [0] O. mykiss TPT1 gene for tumor protein Ots 113242-216 1 exon 3’UTR NM 001165318 [6.49E-153] Salmo salar exostoses (multiple) 1c (ext1c) Ots 113457-40 1 intron n.t. (pseudogene) BT071894 [3.63E-85] Salmo salar C-ets-2 Ots 115987-325 1 intron Ots 117043-255 1 intron Ots 117138-545 1 exon Ots 117242-136 1 exon Ots 117259-271 1 exon 3’ of gene HM159472 [2.14E-41] Salmo salar Foxl2-like protein (Foxl2) gene Ots 117370-471 1 intron CDS AJ003200 [1.97E-55] Danio rerio mRNA for ETS-domain transcription factor PEA3 Ots 117432-409 1 exon Ots 118175-479 1 exon Ots 118205-61 1 exon CDS [Ala>Thr] BT125411 [0] Salmo salar Sjoegren syndrome autoantigen 1 homolog Ots 118938-325 1 intron Ots 120950-417 1 exon Ots 122414-56 1 intron n.t. (pseudogene) BT072655 [2.71E-77] Salmo salar histone deacetylase complex subunit SAP130 Ots 123048-521 1 intron CDS BT149991 [7.43E-56] Salmo salar ribosomal protein S26 mRNA 39 Ots 123205-61 1 exon 3’UTR NM 001139756 [6.84E-38] Salmo salar transposase-like (LOC100194703) Ots 123921-111 1 intron Ots 124774-477 1 exon 3’UTR NM 001173779 [2.01E-165] Salmo salar cAMP-responsive element-binding protein (cr3l2) Ots 126619-400 1 intron 5’ of gene NM 001139980 [1.14E-68] Salmo salar Acyl-CoA desaturase (acod) Ots 127236-62 1 exon Ots 127760-569 1 exon CDS [Ile] NM 001140111 [6.59E-101] Salmo salar Zinc ﬁnger protein 503 (zn503) Ots 128302-57 1 exon 3’UTR BT073572 [5.14E-138] O. mykiss ribosomal protein L20 Ots 128495b-45 1 intron Ots 128693-461 1 intron CDS NM 001141075 [3.57E-94] Salmo salar stathmin-like 4 (stmn4) Ots 128757-61 1 exon 3’UTR BT057575 [3.08E-176] Salmo salar thymosin beta-11 Ots 129144-472 1 exon Ots 129170-683 1 intron Ots 129303b-54 1 intron Ots 129458-451 1 exon 3’UTR NM 001139945 [7.48E-99] Salmo salar FK506 binding protein 8 (fkbp8) Ots 129870-55 1 exon Ots 130720-99 1 exon Ots 131460-584 1 intron CDS BT072067 [5.46E-74] Salmo salar neural cell adhesion molecule L1-like precursor Ots 131802-393 1 exon Ots 131906-141 1 intron n.t. AB258536 [1.62E-49] O. mykiss Onmy-LDA gene for MHC class I antigen Ots AldB1-122 1 unk. 5’ of gene NM 001123627 [3.28E-17] Salmo salar aldolase b and fructose-bisphosphate (aldob) Ots AldoB4-183 1 exon CDS [poor match] NM 001123627 [manual] Salmo salar aldolase b (aldob) Continued on next page Table 1.4 – continued from previous page

Assay name Ref SNP Genic location BLAST # [E-value] Description Ots CathD-141 1 exon 3’UTR (STS) NM 001124711 [1.27E-170] O. mykiss cathepsin D Ots CRB-211 1 intron CDS AF100933 [0] O. mykiss carbonyl reductase dehydrogenase A Ots EndoRB1-486 1 unk. n.t. CU041379 [2.65E-36] Danio rerio DNA sequence in linkage group 16 Ots Hsp90a 1 unk. Originally isolated from heat shock protein 90A Ots Myc-366 1 exon CDS [Asp] NM 001124699 [2.45E-122] O. mykiss v-myc oncogene homolog Ots ALDBINT1-SNP1 1 exon 5’UTR NM 001123627 [1.11E-66] Salmo salar aldolase b (aldob) Ots DESMIN19-SNP1 1 unk. n.t. CR708193 [1.84E-14] Tetraodon nigroviridis full-length cDNA Ots NAML12-SNP1 1 unk. Ots NAML12-SNP2 1 unk. Ots BMP2-SNP1 1 exon CDS [Ser] NM 001173834 [0] Salmo salar bone morphogenetic protein 2 (bmp2) Ots MTA-SNP1 1 intron CDS DQ139342 [5.84E-101] O. tshawytscha metallothionein A (metA) Ots TF1-SNP1 1 exon CDS [Lys>Asn] AF488833S1 [1.26E-147] Salmo salar isolate Ssa-1 transferrin gene Ots ARNT-195 n/a exon 3’UTR NM 001124710 [0] O. mykiss aryl hydrocarbon receptor translocator Ots RAG3 n/a exon 3’UTR OMU73750 [0] O. mykiss RAG gene and 3’UTR Ots AsnRS-60 2 exon CDS [Pro] DQ025751 [3.71E-159] O. tshawytscha Ots.AsnRS.82.35 genomic Ots aspat-196 3 intron CDS EF042601 [0] O. tshawytscha aspartate aminotransferase gene

40 Ots CD59-2 n/a exon 5’UTR NM 001124422 [1.38E-49] O. mykiss CD59-like protein 2 (cd59-2) Ots CD63 n/a exon CDS [Gly] NM 001124496 [1.78E-100] O. mykiss Cd63 antigen (cd63) Ots EP-529 n/a exon CDS [Asp>Asn] NM 001124693 [0] O. mykiss ependymin (om-i) Ots GDH-81x 3 exon 3’UTR EF042600 [0] O. tshawytscha glutamate dehydrogenase gene Ots HSP90B-385 2 exon CDS [Gly] DQ908921 [0] O. tshawytscha heat shock 90 kDa protein gene Ots MHC1 4 exon CDS [Trp>Arg] AF104585 [2.1E-110] O. tshawytsha isolate Onts-B*3 MHC class I alpha 2 Ots mybp-85 n/a unk. Ots myoD-364 3 exon 3’UTR EF042596 [0] O. tshawytscha myoD gene Ots Ots311-101x 3 intron in MSAT repeat AF393194 [7.7E-84] O. tshawytscha msat OtsG311 sequence Ots PGK-54 n/a exon CDS [Lys>Met] NM 001139794 [7.07E-71] Salmo salar Phosphoglycerate kinase (pgk) Ots Prl2 4 intron CDS S66606 [0] O. tshawytscha prolactin II Ots RFC2-558 2 unk. n.t. DQ025583 [6.37E-90] O. tshawytscha clone Ots.RFC2.38.66 Ots SClkF2R2-135 2 intron CDS DQ780892 [4.74E-112] O. tshawytscha CLOCK1a (Clock1a) Ots SWS1op-182 2 intron CDS NM 001124321 [2.08E-98] O. mykiss SWS1 opsin (LOC100135983) Ots TAPBP n/a exon CDS [Ser] NM 001124553 [6.37E-60] O. mykiss tapasin long form Ots u07-07.161 n/a unk. n.t. EU616651 [0] O. tshawytscha isolate u07-07 SNP assay target Ots u07-49.290 n/a unk. n.t. EU616659 [0] O. tshawytscha isolate u07-49 SNP assay target Ots u4-92 2 unk. n.t. DQ025560 [8.66E-176] O. tshawytscha clone Ots.u4.51.26 Ots S71-336 n/a unk. Ots unk-526 n/a intron n.t. GU817335 [1.35E-57] Salmo salar clone BAC CHORI214-083H23 1.5 Discussion

We describe a large set of new genetic resources for Chinook salmon, one of the worlds most economically important fish species and a major component of north Pacific ocean fisheries. A large EST sequencing effort was undertaken that evaluated variation in 225 gene fragments, and in over 131 kb of genomic sequence, in an average of 16 individual salmon each. The resulting 117 SNP assays that were successfully validated in Chinook salmon were broadly polymorphic and had substantial power for biological inference. This effort more than doubles the number of published SNP assays available for this species. Applications for these markers include genetic stock identification (GSI) for mixed fishery and ecological applications, individual identification, linkage mapping and pedigree construction.

GSI is becoming a major component of salmon fishery management, with mixed ocean and inland fisheries being evaluated with molecular markers in all Pacific salmon species. GSI requires markers with frequency differences between populations in a reference baseline database (Seeb et al. 2007). The 117 SNPs described here have a

mean estimated FST of 0.107 for the ﬁve Chinook salmon stocks evaluated here, which

represent the three largest river systems in the coterminous United States: the Sacra-

mento, Klamath and Columbia Rivers. Pairwise FST values for individual loci ranged

from 0 to 0.592, indicating that some of these loci are subject to dramatically diﬀerent

evolutionary forces than others. Since a substantial fraction of these novel SNPs are

located within coding regions of genes, it is possible that some of the polymorphisms

41 targeted by our assays are directly influenced by natural selection. More likely is that a number of these markers are located in genomic regions that have been affected by recent natural selection and the allele frequency differences are the result of hitchhiking effects (Barton 2000). Regardless, the substantial allele frequency differences between populations present at many of these SNP loci indicate that they will be very useful for stock discrimination, exceeding microsatellite loci in discriminatory power when weighted by the total number of alleles. Initial analyses (data not shown) indicate that by selecting the best set of 96 loci, which is convenient for our genotyping platform, from this and other published markers (Smith et al. 2005b; Campbell et al. 2008), assignment accuracy for GSI is as good or better with 96 SNPs than with the 13 standardized microsatellites currently in use (Seeb et al. 2007), for Chinook salmon ocean fishery mixtures south of the Columbia River basin.

Pedigree reconstruction, in the form of large-scale parentage inference, has been proposed as an alternative to physical tags for salmonids and other ﬁshes (Hankin et al.

2005; Garza and Anderson 2007). This method, termed parentage-based tagging (PBT), involves genotyping reproducing individuals and using their genotypes as intergenerational genetic tags that are recovered through parentage inference with their progeny.

PBT has some distinct advantages over traditional large-scale tagging programs such as coded-wire tags, including 1) individual-specific tag recoveries, 2) no tagging or handling of juvenile fish, with their associated very low recovery rates (<2 recoveries per 1000 tags in Chinook salmon; Hankin et al. 2005), 3) fish can be non-lethally sampled during seaward migration, in fisheries, and upon return to spawn, and 4) valuable corollary

42 data in the form of a large number of pedigrees (Garza and Anderson 2007). Over time, some of these pedigrees will become extensive and can serve as the basis for detailed linkage maps and associated mapping of quantitative trait loci (Boulding et al. 2008;

Moen et al. 2008; Pemberton 2008). Goverment agencies have traditionally mitigated the terrestrial and aquatic ecosystem impacts responsible for salmonid population declines with production of ﬁsh in hatcheries and subsequent population supplementation.

Millions of Chinook salmon originate in hatcheries each year and they are the majority of fish in some populations (Barnett-Johnson et al. 2007). Such genetic tagging, and the analysis of the associated pedigrees, could have considerable importance in under- standing the effects of hatchery practices on life history parameters and fitness, since the entire production can be tracked by simply collecting genotypes from all broodstock at spawning.

PBT involves the identiﬁcation of true parents from among very large sets of potential parents, which in turn requires the accurate evaluation of exceedingly small error rates to avoid biologically important rates of false positive pedigree reconstruction.

Anderson and Garza (2006) describe novel importance sampling methods for estimating such probabilities and demonstrated that PBT can be used to accurately reconstruct parent/oﬀspring trios in salmon using 80-100 SNP markers with a mean MAF of 0.20.

The top 96 SNP loci described here have mean MAF of 0.22 (Klamath) to 0.28 (Feather) in the ﬁve focal populations, and the inclusion of other published SNP markers in an optimal set of 96 loci increases mean MAF even further for all populations (data not shown). With pedigree-based inference, minimizing genotyping errors is also critical,

43 since they can cause apparent Mendelian incompatibilities. The lower genotyping error and mutation rates of SNP markers combine to make them the preferred type of marker for the large scale data generation and parentage analyses necessary to implement PBT

(Anderson and Garza 2006; Garza and Anderson 2007). Microsatellite markers are still useful for both GSI and small-scale pedigree reconstruction, particularly for inferring non-parent/offspring relationships, but the reduced staff time and other costs associated with SNP genotype generation, as well as the portability of SNP data, led Hankin et al. (2005) to recommend a transition to SNP markers for multilateral, collaborative research and management of Pacific salmon ocean fisheries.

The MAF requirements for SNP loci in GSI and PBT applications are different. GSI requires frequency differences between populations, which are maximized when loci are fixed for alternative alleles in different populations or lineages. In contrast, power for PBT is entirely dependent upon the mean MAF of the set of SNP loci employed, which is maximized when all loci have two alleles at equal frequency in the focal population. With our balanced ascertainment panel, we were able to discover a set of SNP loci with both MAFs in our focal California populations that exceed those necessary for PBT applications (Anderson and Garza 2006) and also with sufficient allele frequency differences to have high power with GSI. The eventual combination of GSI and PBT analyses on the same genotypic data in a single analytical framework will be a major advance in genetic tagging methodology. With such an integrated GSI/PBT system, all fish genotyped with the same set of markers will yield biological inference, either individual identification when parents are sampled (or a fish is recaptured), or

44 population assignment using a baseline reference database if they are not directly linked to other sampled individuals in a pedigree.

Our discovery eﬀort employed a balanced ascertainment approach, which included an ascertainment sample with representatives of a number of Chinook salmon lineages, and a design criterion that targeted all loci with suﬃcient variation and generally did not discriminate on the basis of the population in which the variation was found.

This strategy led to a set of loci that were similarly variable in all of the populations for which validation was pursued. The inclusion of fish in the ascertainment and validation samples that display variation in migration (yearling and sub-yearling outmigrants) and maturation strategies (fall-run and spring-run types), may also provide additional power for the discrimination of fish from stocks that are differentiated primarily due to these life history strategies (e.g. Central Valley-Fall vs. Central Valley-Spring). Nevertheless, these markers are an upwardly biased sample of the SNP MAF spectrum in the ascertainment populations, because of the three genotypes design criterion, and rare SNP alleles are underrepresented.

It is also important to note that our five validation populations are part of only one of the major lineages of Chinook salmon (Waples et al. 2004; Seeb et al. 2007), with the species extending across the North Pacific rim to Asia, and extensive differentiation throughout the range. However, the ascertainment sample also included representatives of populations from British Columbia lineages, so the SNP markers described here should be more broadly useful in the southeastern part of the species range. Still, these markers are expected to overestimate the mean MAF and proportion of polymorphic

45 loci in populations that are not part of the lineages in the ascertainment sample. Bias corrections can be used to better approximate marker polymorphism and diﬀerentiation for phylogenetically distance populations (Clark et al. 2005; Albrechtsen et al. 2010), but ultimately more SNP markers will need to be ascertained for applications in these

Chinook salmon lineages.

Next-generation sequencing (NGS; pyrosequencing, single base extension) is a potentially powerful method for discovering SNPs and other genomic variation. In- deed, NGS is unparalleled for the identiﬁcation of many potential candidate markers rapidly and at minimal expense. However, when genomic resources and sequence data for a target species exist, the most important components of SNP discovery become

1) validation of observed substitutions as true SNPs and not artifacts, 2) choosing an optimal set of polymorphisms for downstream applications using MAF and linkage criteria, and 3) avoiding ascertainment bias. In highly structured species, such as most salmonid ﬁshes, it is critical to obtain sequence data from the same genomic regions in a diverse sample of individuals to minimize ascertainment bias. There is abundant existing genomic sequence in Oncorhynchus species for SNP discovery, we investigated only about 500 of the nearly 100K EST sequences in the O. mykiss Gene Index (and many of the Salmo salar Gene Index ESTs are likely informative as well), and about 80% of all the gene fragments we investigated had observed substitutions. Since it is easier to ensure identical genomic coverage across individuals with traditional Sanger sequencing than with NGS, and candidate SNPs discovered with NGS are typically resequenced in ascertainment panels prior to assay development anyway, it may be more economical

46 to employ traditional sequencing strategies to develop additional markers for population genetic applications in other parts of the Chinook salmon range. In contrast, the genomic needs for other applications, such as construction of microarray, linkage and physical maps, are likely to be best met with NGS strategies.

47 Chapter 2

Evaluation of a SNP baseline for genetic stock identiﬁcation of Chinook salmon

(Oncorhynchus tshawytscha) in the

California Current Large Marine

Ecosystem1

2.1 Abstract

Chinook salmon from the West Coast of North America are an economically and ecologically important species and a major component of North Paciﬁc Ocean ﬁsh- eries. Their anadromous life history strategy generates populations (or stocks) that are

1accepted with revisions, resubmitted, awaiting editorial approval: Clemento, A.J., E.D. Crandall, J.C. Garza and E.C. Anderson, Fishery Bulletin, 2014

48 frequently genetically differentiated from one another, although not visually discern- able. In many cases, it is desirable to discern the stock of origin of an individual fish or the stock composition of a mixed sample to monitor stock-specific impacts and alter management accordingly. Genetic stock identification (GSI) provides such discrimination and we describe here a novel GSI baseline composed of genotypes from over 8,000 individual fish from 69 distinct populations at 96 single nucleotide polymorphism (SNP) loci. The populations included in the baseline represent the likely sources for over 99% of the fish encountered in ocean salmon fisheries off California and Oregon. This new genetic baseline permits GSI using rapid and cost effective SNP genotyping, and power analyses indicate that it has near maximum power for discriminating most Chinook salmon stocks to the level of resolution needed for fishery management by the Pacific

Fishery Management Council. In an ocean ﬁshery sample, GSI assignments of over 1000

fish, using our baseline, were highly concordant (∼99%) at the reporting unit level to identifications from the physical coded wire tags recovered from the same fish. This

SNP baseline represents an important advance in the technologies available for ﬁshery management and ecological investigation of Chinook salmon at the southern end of their geographic range.

49 2.2 Acknowledgments

The authors would like to thank the entire Molecular Ecology and Genetic

Analysis Team in the Fisheries Ecology Division of the SWFSC for their invaluable assistance with genotyping and analyses. Of critical importance to the successful completion of this project were the baseline samples provided to us by: California Department of Fish and Game (S. Harris), Hoopa Valley Tribal Fisheries Department (G. Kautsky),

Oregon Department of Fish and Wildlife, Oregon State University Department of Fish- eries and Wildlife (M. Banks), Idaho Department of Fish and Game (M. Campbell),

Columbia River Inter-Tribal Fish Commission (S. Narum), NOAA Northwest Fisheries

Science Center (P. Moran), U.S. Fish and Wildlife Service (M. Brown, D. Hawkins, and C. Smith), Washington Department of Fish and Wildlife (S. Blankenship and K.

Warheit), University of Washington School of Aquatic and Fishery Sciences (L. Seeb),

Department of Fisheries and Oceans, Canada (T. Beacham), and Alaska Department of

Fish and Game (W. Templin). Fishery samples were collected by the California Depart- ment of Fish and Game, and provided to us by M. Heisdorf and M. Palmer-Zwahlen.

We also thank T. Beacham and two anonymous referees for comments, which improved this manuscript. This project received funding from NOAAs Cooperative Fisheries Re- search Program and the Southwest Fisheries Science Center. A. Clemento also received support from a California Bay Delta Science Fellowship and the University of California

Coastal Environmental Quality Initiative. Many of the baseline samples were collected and DNA extracted with funds from the Paciﬁc Salmon Commission.

50 2.3 Introduction

Chinook salmon (Oncorhynchus tshawytscha) are found in rivers from central

California around the North Paciﬁc Rim to Russia (as well as those draining into the

Bering Sea), and are the target of valuable commercial and recreational fisheries. A key component of the Chinook salmon life history is natal homing, whereby these anadromous fish typically return to spawn in the same river in which they were born. This homing generates populations (or stocks) that may be genetically differentiated from neighboring populations and can exhibit local adaption (Utter 1989, Taylor 1991). Re- cent population declines, particularly at the southern end of the species native range where many stocks are listed under the US Endangered Species Act (ESA; Federal

Register 1990, 1999), have highlighted the need to refine the management and conservation of Chinook salmon. However, such refinements are challenging, because the migratory life history of salmon means that the many anthropogenic impacts occurring in rivers or in the ocean (e.g. fisheries, water diversion, or turbine entrainment) may affect multiple, intermingled stocks. In such cases, it may be necessary to discern the stock of origin of affected fish to monitor stock-specific impacts and design management strategies accordingly.

The use of pre-existing biological markers to distinguish salmon stocks has a long history. The traits used in these eﬀorts have included morphometric and meristic characters (Fournier et al. 1984, Claytor and MacCrimmon 1987), scale patterns (Cook,

1982), parasite assemblages (Boyce 1985), and stable isotope ratios (Barnett-Johnson

51 et al. 2008). However, the most universally applicable methods have involved the use of genetic markers, since every fish has a unique genetic makeup. The first genetic markers widely used for identification in salmon were electrophoretically detectable protein polymorphisms known as allozymes (Milner et al. 1985, Shaklee and Phelps 1990,

Tessier et al. 1995, Allendorf and Seeb 2000). With the advent of the polymerase chain

reaction (PCR), many more types of genetic markers became available to discriminate

salmon populations, including mitochondrial DNA polymorphisms (Cronin et al. 1993),

minisatellites (Beacham et al. 1996, Miller et al. 1996), microsatellites (Seeb et al. 2007,

Moran et al. 2013), ampliﬁed fragment length polymorphisms (Flannery et al. 2007)

and, most recently, single nucleotide polymorphisms (SNPs; Smith et al. 2005a, Smith

et al. 2005b, Aguilar and Garza 2008, Narum et al. 2008, Abad´ıa-Cardoso et al. 2011,

Clemento et al. 2011).

Genetic stock identiﬁcation (GSI) typically proceeds in two steps. First, sam-

ples are collected from potential source populations and genotyped with a set of genetic

markers in order to estimate population allele frequencies. These genotypes are called

the baseline. Then, data from individuals sampled from a mixed-stock collection (a mix-

ture) and genotyped with the same set of genetic markers are compared to the baseline

to estimate the relative proportions of individuals from each of the represented source

populations. Single individuals of unknown origin can also be assigned to speciﬁc pop-

ulations. GSI inference is typically carried out using maximum likelihood or Bayesian

methods (Smouse et al. 1990, Pella and Masuda 2000).

The ﬁrst large-scale baseline for GSI of Chinook salmon utilized allozyme mark-

52 ers (Teel et al. 1999), but technical and logistical issues limited their future appeal. The allozyme database was supplanted in Canada by a microsatellite baseline developed by the Department of Fisheries and Oceans (Beacham et al. 2006), and more broadly by a microsatellite baseline database developed through a large, international collaboration

(Seeb et al. 2007). This collaboration required enormous effort to standardize data across labs, as microsatellite allele names and sizes are not usually consistent between different labs and genotyping equipment. The Seeb et al. (2007) microsatellite baseline has been an effective tool for GSI but has a number of disadvantages: genotyping and scoring of microsatellites is labor-intensive; genotyping error rates can be relatively high, making the 13 microsatellites in that baseline inadequate for applications such as pedigree reconstruction (Anderson and Garza 2006, Garza and Anderson 2007, Abad´ıa-

Cardoso et al. 2013); missing data rates can also be quite high; and, ﬁnally, any new laboratory that wishes to use the baseline must undertake a costly standardization process. Additionally, it has now been demonstrated that SNPs, despite typically having only two alleles per-locus, do have suﬃcient power to be successfully employed in a GSI context with a modest number of genetic markers (Smith et al. 2007, Narum et al. 2008,

Templin et al. 2011, Larson et al. 2013).

Early simulation studies suggested that the biallelic nature of SNPs would make them less useful than highly polymorphic microsatellites for population discrimination (Bernatchez and Duchesne 2000, Kalinowski 2004). However, SNPs are located throughout the genome and may be discovered in genetic regions with higher than average divergence (Nosil et al. 2009), increasing their utility for GSI. Moreover, SNPs do

53 not suﬀer from many of the disadvantages of microsatellites: SNP markers are amenable to the automated, high-throughput genotyping required for large projects; SNP genotyping error rates are very low, making them suitable for pedigree reconstruction; and, importantly, SNP assays do not typically require standardization between labs, so a

SNP baseline is immediately useful to any group or agency that genotypes a mixture sample with the markers used in the baseline (Seeb et al. 2011).

Here, we describe the development and evaluation of a new baseline of SNP marker data for Chinook salmon in the southern part of their native range for use in ecological investigation in the California Current Large Marine Ecosystem (and its tributaries) and in ﬁsheries managed by the Paciﬁc Fishery Management Council (PFMC).

We introduce a panel of 96 SNP markers and a baseline of nearly 8,000 salmon from 68

Chinook salmon populations ranging from California to Alaska. We describe the procedures used to select these SNP markers from amongst a larger number of candidates and document the resulting patterns of genetic diﬀerentiation between various populations.

We evaluate the power of the new baseline for GSI by both self-assignment and simulated mixture analyses, focusing on stocks commonly encountered in PFMC ﬁsheries.

Finally, we analyze 2,090 fish sampled in 2010 from the sport and commercial fisheries off the coast of California and compare the results of these analyses to the coded wire tag (CWT) data from these fish to demonstrate the effectiveness of the baseline for classifying individuals to specific management units.

54 2.4 Methods

2.4.1 Baseline Populations

Populations were selected for inclusion in the baseline to provide broad geographic coverage across the range of Chinook salmon in the coterminous United States, from Washington to California, while also allowing for the identification of fish from elsewhere in the species’ geographic range. Adult fish were sampled on spawning grounds, in terminal fisheries or at hatcheries over the last decade and were provided by numerous contributors (see Acknowledgments and Warheit et al. 2013). We included populations expected to be encountered in ocean fisheries off California and Oregon, as well as populations with special management status (e.g. ESA-listed). Accordingly, the major lineages of Chinook from California and Oregon are emphasized in the baseline, as were populations distinguished by life history strategy (spring-run, fall-run, winter-run, etc.), but representatives of the major lineages from further north were also included. DNA was extracted from samples for California populations using DNEasy Blood and Tissue kits on a BioRobot3000 (QIAGEN, Inc., Valencia, CA) according to the manufacturers protocols, while DNA from populations in Oregon, Washington, Canada and Alaska was extracted by the contributors (see Acknowledgments) using various methods. Sample sizes ranged from 44 to 1409 individuals per population and averaged 116. The 1409

ﬁsh from the Trinity River Hatchery were initially genotyped with our SNP panel for another purpose, but were included here in total to provide a comprehensive reference sample for identifying this important group. Excluding this disproportionately large

55 sample, the average number of individuals per population was 97. In total, the baseline includes 7,984 Chinook salmon from 68 distinct populations (Table 2.1).

Each population in the baseline belongs to a single reporting unit, a designation established in previous GSI work that reﬂects a combination of “genetic similarity, geographic features and management applications” (Seeb et al. 2007). Reporting units are generally composed of multiple populations that share genetic similarity or are subject to similar management regimes. The 68 Chinook salmon populations in our baseline fall into 38 distinct reporting units (Table 2.1) and some reporting units in

Alaska and Canada are represented by only a single population.

Coho salmon (Oncorhynchus kisutch) are occasionally misidentified as Chinook salmon in ocean fisheries and in ecological sampling. We included a collection of 47 coho salmon from California as the 69th population in the baseline to assist in identifying coho salmon that have been incorrectly identified as Chinook salmon.

56 Table 2.1: Populations and reporting groups in the single-nucleotide polymorphism baseline for genetic stock identiﬁcation of Chinook salmon from the West Coast of North America. Shown are the names used on the phylogeographic tree (Figure 2.1), the total number of individuals sampled (n), the number used in the training set (nt), estimates of unbiased expected (Exp.) and observed (Obs.) heterozygosity (Hz), and the mean number of alleles (A); also shown are the proportion of individuals that self-assign (Assign.) to the population (pop.) from which they were sampled and the proportion that self-assign to the correct reporting (rep.) group, as well as the mean FST for each population within and between reporting groups. Note that mean summary values shown were calculated excluding the coho salmon sample.

Assign. Mean Mean

Exp. Obs. Assign. to rep. FST FST Reporting Group Population Tree Name n nt Hz Hz A to pop. group w/in btw. Central Valley spring Butte Creek spring CVsp Butte 425 26 0.357 0.33 1.99 0.68 0.93 0.017 0.196 Mill Creek spring CVsp Mill 145 23 0.371 0.377 1.99 0.48 0.8 0.012 0.173 Deer Creek spring CVsp Deer 119 12 0.367 0.346 1.99 0.5 0.8 0.013 0.174 Up. Sacramento R. sp. CVsp UpSac late 372 0.368 0.355 1.99 0.26 0.78 0.008 0.175 Central Valley fall Feather R. Hatchery sp. CVfl FeatherRHsp 470 47 0.373 0.374 1.99 0.44 0.87 0.009 0.179 57 Feather R. Hatchery fall CVfl FeatherRHfl 146 23 0.37 0.371 1.98 0.18 0.85 0.004 0.19 Butte Creek fall CVfl Butte 188 0.369 0.355 2 0.13 0.91 0.003 0.187 Mill Creek fall CVfl Mill 97 12 0.366 0.358 1.98 0.14 0.95 0.004 0.2 Deer Creek fall CVfl Deer 70 0.363 0.347 1.98 0.29 0.9 0.005 0.195 Mokelumne River fall CVfl Mklmne 95 27 0.37 0.373 1.98 0.26 0.94 0.005 0.198 Battle Creek fall CVfl Battle 141 23 0.369 0.351 1.99 0.29 0.89 0.005 0.188 Up. Sac. R. late-fall CVfl UpSac 93 23 0.367 0.364 2 0.54 0.93 0.01 0.193 Central Valley winter Sacramento R. winter Sac win 295 19 0.297 0.289 1.97 1 1 - 0.263 California Coast Eel River CACoast Eel 95 12 0.327 0.321 2 0.89 0.96 0.029 0.203 Russian River CACoast Russian 94 0.368 0.372 2 0.84 0.98 0.029 0.156 Klamath River Iron Gate Hatchery Klamath IronGH 117 12 0.326 0.345 1.97 0.97 0.99 0.053 0.232 Trinity River Hatchery Klamath TrinityHsp 1409 12 0.318 0.312 2 0.93 0.97 0.053 0.243 N. California/ Smith River nCal sOR Smith 159 0.377 0.381 1.99 0.77 0.87 0.014 0.138 S. Oregon Coast Chetco River nCal sOR Chetco 94 11 0.372 0.367 1.99 0.73 0.86 0.014 0.137 Rogue River Cole Rivers Hatchery Rogue ColeRHsp 141 11 0.367 0.362 2 0.62 0.86 0.006 0.155 Applegate Creek Rogue Applgt 92 0.369 0.361 2 0.5 0.77 0.006 0.153 Mid Oregon Coast Coquille River mOR Coquille 47 0.352 0.343 1.99 0.72 0.83 0.039 0.151 Umpqua River spring mOR Umpqua 137 11 0.386 0.375 2 0.63 0.64 0.055 0.119 Siuslaw River mOR Siuslaw 93 0.345 0.348 1.98 0.4 0.46 0.04 0.146 Continued on next page Table 2.1 – continued from previous page

Assign. Mean Mean

Exp. Obs. Assign. to rep. FST FST Reporting Group Population Tree Name n nt Hz Hz A to pop. group w/in btw.

North Oregon Coast Nestucca Hatchery nOR NestuccaH 48 0.338 0.328 1.96 0.71 0.83 0.029 0.16 Alsea River nOR Alsea 131 0.335 0.309 2 0.47 0.76 0.042 0.159 Nehalem River nOR Nehalem 93 0.316 0.317 1.96 0.97 0.99 0.059 0.193 Siletz River nOR Siletz 93 0.331 0.33 1.98 0.69 0.81 0.031 0.163 Willamette River N. Santiam Hatchery Will NSantiamH 93 0.324 0.327 1.95 0.8 0.99 0.014 0.181 McKenzie Hatchery Will McKenzHsp 48 0.334 0.376 1.94 0.69 0.96 0.014 0.175 Deschutes River fall Lower Deschutes River Deschutes ﬂ 94 0.366 0.357 2 0.56 0.56 - 0.145 Low. Columbia R. fall Cowlitz Hatchery fall COlow CowHﬂ 141 0.365 0.374 1.99 0.79 0.79 - 0.142 Lower Columbia R. Cowlitz Hatchery sp. COlow CowHsp 44 11 0.368 0.37 1.97 0.67 0.7 0.029 0.16 spring Kalama Hatchery sp. COlow KalamaHsp 48 12 0.372 0.359 1.99 0.5 0.61 0.029 0.134 Mid Columbia R. tule Spring Creek Hatchery COmid SpringCH 142 0.322 0.331 1.97 0.97 0.97 - 0.206 Upper Columbia R. Hanford Reach COup Hanford 92 0.355 0.353 1.99 0.36 0.76 0.002 0.166

58 summer/fall Priest Rapids Hatchery COup PriestHsumfl 48 0.361 0.359 1.99 0.25 0.83 0.002 0.164 Wells Hatchery COup WellsHsumfl 48 0.355 0.369 1.99 0.46 0.92 0.004 0.175 Mid/Up. Columbia Wenatchee River COmup Wenatchee 48 0.209 0.202 1.88 0.85 0.85 0.048 0.26 River spring Cle Elum Hatchery COmup CleEHsp 48 0.262 0.255 1.95 0.94 0.96 0.048 0.219 Snake River fall Lyons Ferry Hatchery Snake LyonsFHfl 119 12 0.359 0.36 2 0.45 0.45 - 0.158 Snake River Rapid River Hatchery Snake RapRHsumsp 48 0.191 0.194 1.84 0.85 0.94 0.034 0.272 spring/summer McCall Hatchery Snake MCallHsumsp 48 0.199 0.196 1.84 0.75 0.96 0.034 0.278 Washington Coast Forks Creek Hatchery WACoast ForksCH 93 0.35 0.345 1.99 0.89 0.94 0.042 0.143 Quinalt Lake fall WACoast Quinalt 48 0.348 0.341 1.98 0.9 0.96 0.042 0.152 South Puget Sound Soos Creek Hatchery sPuget SoosCH 142 0.36 0.358 2 0.91 0.91 - 0.158 North Puget Sound Kendall Hatchery sp. nPuget KendlHsp 48 0.326 0.336 1.95 0.92 0.96 0.042 0.17 Marblemount H. sp. nPuget MrblHsp 48 0.343 0.337 1.99 0.92 0.94 0.042 0.156 Lower Fraser River Harrison River Fraser Harris 48 0.329 0.326 1.98 0.96 0.96 0.152 0.169 Birkenhead Hatchery Fraser BirkenH 91 0.259 0.255 1.84 1 1 0.152 0.231 Lower Thompson R. Spius Creek Hatchery Thompson SpiusCH 46 11 0.271 0.275 1.89 1 1 - 0.201 Eastern Vancouver I. Big Qualicum Hatchery eVancI BigQual 48 0.352 0.338 2 0.83 0.83 - 0.145 Western Vancouver I. Robertson Hatchery wVancI RobHfl 48 0.341 0.364 1.98 0.96 0.96 - 0.152

Continued on next page Table 2.1 – continued from previous page

Assign. Mean Mean

Exp. Obs. Assign. to rep. FST FST Reporting Group Population Tree Name n nt Hz Hz A to pop. group w/in btw. Lower Skeena River Lower Kalum River lSkeena Kalum 48 0.303 0.303 1.96 0.77 0.77 - 0.156 Mid Skeena River Morice River mSkeena Morice 47 0.279 0.276 1.89 0.81 0.91 0.013 0.173 Kitwanga River mSkeena Kitwanga 48 0.291 0.29 1.94 0.54 0.75 0.013 0.175 S. Southeast AK LPW - Unuk R. stock sSEAK Unuk 48 0.301 0.29 1.94 0.79 0.79 - 0.165 Gulf AK Alsek R. Goat Creek AlsekAK Goat 48 0.243 0.245 1.69 0.96 0.96 - 0.248 Gulf AK Karluk R. Karluk River KarlukAK 47 0.23 0.22 1.73 1 1 - 0.237 Taku River L. Tatsamenie Lake Taku LilTats 48 0.271 0.265 1.92 0.9 0.9 - 0.188 NSE AK Chilkat R. Pullen Creek Hatchery nSEAK PullenCH 48 0.26 0.276 1.77 0.98 0.98 - 0.209 Gulf AK Situk R. Situk River SitukAK 48 12 0.244 0.248 1.77 0.94 0.94 - 0.21 Copper River Sinona Creek CopperAK Sinona 47 0.229 0.226 1.63 0.98 0.98 - 0.244 Susitna River Montana Creek SusitnaAK Montana 48 0.21 0.201 1.73 0.92 0.92 - 0.249

59 Lower Kuskokwim/ George River WestAK George 47 0.234 0.229 1.78 0.43 0.98 0.004 0.239 Western AK Kanektok River WestAK Kanektok 48 0.241 0.232 1.81 0.38 0.96 0.001 0.233 Togiak River WestAK Togiak 48 0.241 0.229 1.79 0.4 0.94 0.005 0.233 Mid Yukon River Kantishna River Yukon Kantishna 48 0.208 0.204 1.67 0.94 0.94 - 0.269 Coho salmon California Coho Coho 47 0.089 0.094 1.33 1 1 - 0.463 total 8031 mean 0.32 0.317 1.93 0.69 0.88 0.028 0.188 2.4.2 Markers and Genotyping

We compiled 192 Taqman c (Life Technologies Corporation, Carlsbad, CA), or

5’ nuclease, SNP genotyping assays from previously published discovery eﬀorts (Smith et al. 2005a, 2005b; Campbell and Narum 2008; Narum et al. 2008; Clemento et al. 2011) to test their scorability and power for GSI. Taqman c technology combines standard

PCR primers targeting the genomic region around a SNP with two different fluorescent probes that identify the two nucleotide bases present at the SNP. Per the manufacturers recommendation, a multiplex pre-amplification reaction was used to increase the copy number of targeted genomic regions. Multiplex PCR products were diluted with 15µL of 2mM Tris and frozen. Samples were then genotyped on 96.96 Dynamic Genotyping

ArraysTM using an EP-1 genotyping system (Fluidigm Corporation, South San Fran- cisco, CA) according to the manufacturers protocols. Fluidigm Dynamic Arrays use integrated ﬂuidic circuitry and PCR volumes of ∼9nL to simultaneously determine the genotype at 96 SNP loci for 96 samples (two of which are no-DNA template controls).

Genotypes were determined using the Fluidigm SNP Genotyping Analysis Software

(version 2.1.1). Genotype determination using quantitative PCR methods involves dis- cerning, on a two dimensional graph, clusters of the ﬂuorescence intensity of the probes for the two alleles; the two homozygote clusters have ﬂuorescence primarily from only one probe, while a heterozygote cluster has similar intensities from both.

60 2.4.3 Marker Selection

A panel of 95 SNP markers was selected from amongst the 192 candidates, reserving one marker for a species identiﬁcation assay (see below). The risk of high grading bias (i.e. wrongly inﬂating the apparent resolving power of a group of loci for

GSI) is particularly great when selecting a panel of markers to distinguish between populations that are closely related, as are many in our baseline. To avoid high grading bias, we employed the Training-Holdout-Leave-One-Out (THL) procedure of Anderson

(2010), which requires that the data be split into training and holdout sets. Training- set genotypes are used to select the loci included in the baseline and can be included in the eventual baseline, but they are not used to evaluate its performance. Rather, performance of the baseline is determined with simulation and self-assignment using only the holdout set, which was not used in any way to select the baseline loci. We chose a training set of 372 individuals drawn from 22 populations (14 from California, three from Oregon, three from Washington, one from British Columbia and one from

Alaska) for initial genotyping with all 192 loci.

For each locus, k, the observed relative frequencies, pik and qik, of the two SNP alleles were calculated for each population, i, in the training set. These values were then used to compute the expected probability of misassignment, P(Misijk), between every

pair of populations i and j using only a single locus i:

2 2 P(Misijk) = 0.5 [ δ(pik≤pjk)pik + δ(pikqik≤pjkqjk)2pikqik + δ(qik ≤qjk)qik +

2 2 δ(pik≥pjk)pjk + δ(pikqik≥pjkqjk)2pjkqjk + δ(qik ≥qjk)qjk ]

61 for all k, where δ(x) = 1 if the condition x is true and 0 otherwise. The values of P(Misijk)

were used to rank the loci for their suitability for resolving between populations i and j

in GSI; lower P(Misijk) implies better resolving power.

The rankings obtained from P(Misijk) were combined with other criteria in a non-automated process to select the final panel of loci (Table 2.2). Each SNP assay was evaluated for scorability and evidence of Hardy-Weinberg (H-W) or linkage disequilibrium. Assays with overly dispersed clusters, more than three clusters, or inadequate spacing between clusters were excluded. Loci with significant deviations from equilibrium expectations were also removed. SNPs with large allele frequency differences between populations are particularly effective for GSI, while SNPs with high minor allele frequencies (MAFs) are most useful for parentage analysis (Anderson and Garza

2006). The remaining 168 loci were then ranked by their MAFs in hatchery populations to be included in pedigree reconstruction studies (see Discussion). Previous simulations indicated that about 100 loci with a MAF greater than 0.2 would be required to achieve the necessary statistical power to assign parentage with suﬃciently low false-negative and false-positive rates (Anderson and Garza 2006). However, the observed MAFs for many loci were in fact greater than 0.2 (and as high as 0.5), meaning that the desired statistical power could be achieved with fewer loci. We therefore selected the 70 loci with the highest MAF in the Feather River population, the primary target for subsequent parentage investigations. The P(Misijk) rankings were then utilized to select 25 additional loci that were useful for distinguishing between diﬃcult-to-resolve populations and reporting units. Finally, an assay to discriminate between Chinook and coho

62 salmon was included as the 96th assay for the 96.96 genotyping arrays.

63 Table 2.2: List of the 96 single nucleotide polymorphism loci used to construct the baseline for genetic stock identiﬁcation of Chinook salmon from the West Coast of North America, including dbSNP accession numbers (at the NCBI on-line repository for short genetic variations) and source reference (SR) where available: 1. Clemento et al. 2011; 2. Smith et al. 2005a; 3. Campbell and Narum 2008; 4. Smith et al. 2005b.; 5. Narum et al. 2008

Locus dbSNP SR Locus dbSNP SR Locus dbSNP SR Ots 94857-232 ss275518685 1 Ots 110495-380 ss275518741 1 Ots 131906-141 ss275518787 1 Ots 96222-525 ss275518688 1 Ots 110551-64 ss275518742 1 Ots AldB1-122 ss275518788 1 Ots 96500-180 ss275518689 1 OkiOts 120255-113 unpubl. - Ots AldoB4-183 ss275518789 1 Ots 97077-179 ss275518691 1 Ots 111312-435 ss275518746 1 Ots Myc-366 ss275518795 1 Ots 99550-204 ss275518695 1 Ots 111666-408 ss275518747 1 Ots ALDBINT1-SNP1 ss275518796 1 Ots 100884-287 ss275518696 1 Ots 111681-657 ss275518748 1 Ots NAML12-SNP1 ss275518798 1 Ots 101119-381 ss275518697 1 Ots 112208-722 ss275518749 1 Ots ARNT-195 unpubl. - Ots 101704-143 ss275518699 1 Ots 112301-43 ss275518750 1 Ots RAG3 n/a 5 Ots 102213-210 ss275518702 1 Ots 112419-131 ss275518751 1 Ots AsnRS-60 ss48398657 2

64 Ots 102414-395 ss275518703 1 Ots 112820-284 ss275518752 1 Ots aspat-196 ss65917744 3 Ots 102420-494 ss275518704 1 Ots 112876-371 ss275518753 1 Ots CD59-2 unpubl. - Ots 102457-132 ss275518705 1 Ots 113242-216 ss275518754 1 Ots CD63 unpubl. - Ots 102801-308 ss275518706 1 Ots 113457-40 ss275518755 1 Ots EP-529 unpubl. - Ots 102867-609 ss275518707 1 Ots 117043-255 ss275518757 1 Ots GDH-81x ss65917741 3 Ots 103041-52 ss275518708 1 Ots 117242-136 ss275518759 1 Ots HSP90B-385 ss65713207 2 Ots 104063-132 ss275518711 1 Ots 117432-409 ss275518762 1 Ots MHC1 ss49851328 4 Ots 104569-86 ss275518714 1 Ots 118175-479 ss275518763 1 Ots mybp-85 unpubl. - Ots 105105-613 ss275518715 1 Ots 118205-61 ss275518764 1 Ots myoD-364 ss65917726 3 Ots 105132-200 ss275518716 1 Ots 118938-325 ss275518765 1 Ots Ots311-101x ss65917748 3 Ots 105401-325 ss275518718 1 Ots 122414-56 ss275518767 1 Ots PGK-54 n/a 5 Ots 105407-117 ss275518719 1 Ots 123048-521 ss275518768 1 Ots Prl2 ss49851322 4 Ots 106499-70 ss275518724 1 Ots 123921-111 ss275518770 1 Ots RFC2-558 ss48398670 2 Ots 106747-239 ss275518725 1 Ots 124774-477 ss275518771 1 Ots SClkF2R2-135 ss48398694 2 Ots 107074-284 ss275518726 1 Ots 127236-62 ss275518773 1 Ots SWS1op-182 ss48398635 2 Ots 107285-93 ss275518728 1 Ots 128302-57 ss275518775 1 Ots TAPBP n/a 5 Continued on next page Table 2.2 – continued from previous page

Locus dbSNP SR Locus dbSNP SR Locus dbSNP SR Ots 107806-821 ss275518730 1 Ots 128693-461 ss275518777 1 Ots u07-07.161 unpubl. - Ots 108007-208 ss275518731 1 Ots 128757-61 ss275518778 1 Ots u07-49.290 unpubl. - Ots 108390-329 ss275518732 1 Ots 129144-472 ss275518779 1 Ots u4-92 ss48398636 2 Ots 108735-302 ss275518733 1 Ots 129170-683 ss275518780 1 Ots BMP2-SNP1 ss275518800 1 Ots 109693-392 ss275518737 1 Ots 129458-451 ss275518782 1 Ots TF1-SNP1 ss275518802 1 Ots 110064-383 ss275518738 1 Ots 130720-99 ss275518784 1 Ots S71-336 n/a 5 Ots 110201-363 ss275518739 1 Ots 131460-584 ss275518785 1 Ots unk 526 n/a 5 65 2.4.4 Population Genetics Analyses

The 7669 samples that were not in the training set for locus selection were genotyped with the ﬁnal panel of 96 SNPs and used as the holdout set in subsequent power analyses (see next section). This holdout set was also used for standard population genetics analyses. We tested each locus-population pair for deviations from

H-W equilibrium using the complete enumeration method (Louis and Dempster 1987) in GENEPOP vers. 4.0 (Rousset 2008). Similarly, in each population, all pairwise locus combinations were investigated for linkage disequilibrium (LD). Default Markov chain parameters were used, except for the number of batches which was increased to 500 to reduce the standard error to acceptable levels (< 0.02; Rousset 2008).

FST was estimated (with θ of Weir and Cockerham 1984) between all pairs of populations using the software package GENETIX vers. 4.05 (Belkhir 1996-2004).

The dataset was permuted 1000 times to determine the signiﬁcance of FST estimates.

Phylogeographic trees were constructed with Cavalli-Sforza and Edwards’ (1967) chord distance (DCE) and the neighbor-joining algorithm in PHYLIP vers. 3.69 (Felsenstein

2005) and were visualized with DENDROSCOPE (Huson et al. 2007). Majority-rule consensus values were calculated from 10,000 bootstrap samples of the data using the

PHYLIP component CONSENSE. The FST values and genetic distances computed should provide an inﬂated estimate of the isolation between populations because the

SNP loci used in the analysis are not a random sample from the genome, as some were chosen for their power in resolving population pairs in our baseline. Nonetheless,

66 these estimates are useful for assessing the relative genetic diﬀerentiation among these populations.

2.4.5 Power Analyses

Three diﬀerent methods were used to assess the power of the SNP baseline for

GSI. First, we performed a self-assignment analysis, and subsequently generated and analyzed simulated mixtures using two diﬀerent procedures.

In self-assignment, allele frequencies for each potential source population are estimated from the samples. Then, for each individual, the probability of its genotype occurring in each population (assuming H-W and linkage equilibria) is calculated, and the individual is assigned to the population for which its genotype probability is highest. We used the likelihood method of Rannala and Mountain (1997), implemented in the software GSI SIM (Anderson et al. 2008), to compute the genotype probabilities, employing a leave-one-out procedure that excludes the gene copies of the individual being assigned and recalculates population allele frequencies prior to assignment. Anal- ogous to the THL procedure of Anderson (2010), both the training and holdout sets were included for estimating population allele frequencies. However, assignments of the training set individuals were excluded from the results to avoid any high-grading bias of assignment accuracy (Anderson 2010).

Analysis of simulated mixed ﬁsheries is a common method for evaluating the resolving power of a baseline for stock identiﬁcation (Fournier et al. 1984, Wood et al.

1987, Kalinowski 2004, Beacham et al. 2006). In many studies, samples from simulated

67 fisheries consisting entirely of fish from one population are analyzed; so called 100% simulations. However, such simulations do not typically assess how well the baseline will perform on samples from fisheries that exploit more than one stock. Therefore, we conducted simulations using 20 different mixing proportion vectors, the population composition of which was constructed by using the baseline to estimate mixing proportions from one of 20 different month-by-area strata from GSI data collected from commercial

fisheries off the coast of California and Oregon in 2010 and 2011 (E. Crandall et al. unpubl. data). These vectors reflect mixing proportions we expect to encounter in PFMC

fisheries. For a given value of the mixing proportion vector of all populations, a replicate simulation consisted of: 1) simulating the number of fish from each population in a sample size of 200 by drawing a multinomial random variable with cell probabilities equal to the mixing proportion vector; 2) simulating the genotypes of the individuals from each population in the mixture sample using two different techniques (cross-validation over gene copies [CV-GC] and K-fold cross-validation [K-Fold], see below); 3) calculating the maximum likelihood estimator (MLE) of the mixture proportions for all the populations from the simulated sample using the baseline, which contains all training and holdout individuals; and 4) estimating the mixing proportion of each reporting unit by summing the mixing proportion estimates of its constituent populations. For each of the 20 values of the mixing proportion vectors, 20,000 replicates were conducted using

CV-GC, while 1,000 replicates were conducted using K-Fold. For both methods, the 5% and 95% quantiles of the distribution of the MLE of reporting unit proportions were calculated from the replicates for each mixing proportion vector.

68 Simulations were undertaken in two diﬀerent ways. With CV-GC, genotypes were simulated by randomly sampling gene copies from the holdout set (to avoid high- grading bias) and those same gene copies were removed from the baseline when calculating the likelihood of population origin for the simulated individual (see Anderson et al. 2008). With K-fold, genotypes were simulated by drawing entire individuals without replacement (jackknife) from the holdout set to form the mixture sample. Those sampled individuals were not included in the baseline, but all unsampled individuals from the holdout set were included in the baseline for estimating the mixing proportions.

2.4.6 Mixed Fishery Samples

Samples from 2,090 salmon landed in ﬁsheries in 2010 were collected by the

California Department of Fish and Game (CDFG) at California ports. Just over half of these ﬁsh carried coded wire tags (CWTs) that identiﬁed their population of origin.

All samples were genotyped with our panel of 96 loci. Individuals successfully genotyped at fewer than 60 loci were removed from further analysis. Failed genotypes were ones that either clustered with negative controls during scoring or fell outside of deﬁned heterozygote and homozygote clusters, likely indicating sample contamination (Smith et al. 2011, Larson et al. 2013). We also used an individual heterozygosity (iHz; the proportion of heterozygous loci for each ﬁsh) criterion of iHz > 0.56 to identify and exclude samples potentially contaminated by DNA from other samples. Simulations of contaminated genotypes using observed allele frequencies indicated little overlap in the distribution of iHz for contaminated and uncontaminated samples (data not shown)

69 and that uncontaminated samples rarely had iHz > 0.56. We used the maximum likelihood framework in GSI SIM to estimate the mixing proportion of different populations amongst the 2,090 fish, and then used that MLE as the prior for calculating the posterior probability of population of origin for each fish. Posterior probabilities of originating from different reporting units were obtained by summing the population-specific probabilities over all populations in a reporting unit. Individuals were then assigned to the reporting unit with the highest posterior probability.

Since all fish will be assigned to a maximum a posteriori (MAP) population regardless of true origin, we employed a simulation method similar to that in Cornuet et al. (1999), but modified to account for missing data, to detect fish that might originate from a population that is not in the baseline, or has an otherwise aberrant genotype.

Briefly, for each fish from the fishery assigned to a population, the allele frequencies from the MAP population were used to simulate 10,000 genotypes with an identical pattern of missing data (if any) as the fish that was assigned. The log-probability of each simulated genotype was computed, given that it came from the population it was simulated from, and then the distribution of those values was compared to the log- probability, La, of the actual assigned fishs genotype, given the allele frequencies in the

MAP population, on the basis of a z-score (La minus the mean of the simulated values, all divided by the standard deviation of the simulated values). The z-score calculation is done conditional on the exact pattern of missing data and is implemented in the C programming language as part of the GSI SIM software. A low-conﬁdence assignment was deﬁned to be one that had a z-score < -3.0 and either a reporting unit posterior

70 probability less than 0.9 or fewer than 90 loci successfully genotyped. Fish with low conﬁdence assignments were left in an unassigned category.

2.5 Results

2.5.1 Genotyping and Basic Population Genetics

We successfully genotyped 8,031 samples from 69 populations for the baseline and submitted the data to the Dryad Digital Repository (http://www.datadryad.org).

All individuals were retained in the baseline, regardless of missing data, as we desired a realistic representation of missing data patterns for subsequent power analyses. One locus failed to amplify entirely in the Copper River population, while three loci failed in the coho salmon sample. Unbiased estimates of heterozygosity (Nei 1978) ranged from

0.194 in the Snake River-Rapid River Hatchery stock to 0.381 in the Smith River population. The coho salmon in the baseline had very low heterozygosity (0.094). Observed heterozygosity and mean number of alleles were generally lower for populations from north of the Columbia River (Table 2.1), likely due to ascertainment bias. Signiﬁcant deviations from HWE (P < 0.0001) were observed at various loci in 17 populations, but represented < 0.3% of all observations. Only the Butte Creek spring-run, Trinity River

Hatchery spring-run and Smith River populations were not in HWE at more than two loci, with five, five and four significant tests, respectively. Similarly, only three loci deviated from HWE in more than two populations: Ots u07 07.161 in three populations,

Ots 111312-435 in six and Ots 111666-408 in four. Only one population (Trinity River

71 Hatchery spring-run) displayed significant LD (P < 0.001) at more than 1% of locus comparisons (1.14%) and, over all populations, the percentage of significant comparisons was 0.16%. Only two locus pairs were significant in more than five populations:

Ots AldB1-122 and Ots AldoB4-183, known to be in same gene complex, were in LD in

42 populations, while Ots Myc-366 and Ots unk-526 displayed LD in eight populations.

A large range in the degree of diﬀerentiation between populations was observed

(Table 2.1). Mean FST across all populations (excluding coho salmon) was 0.183, indicating that approximately 18% of genetic variation was partitioned between population samples. Within reporting units containing more than one population (n = 18), pairwise

FST was between 0.000 and 0.152 with a mean value of 0.018. Ten pairwise comparisons, all within reporting units, were not signiﬁcantly diﬀerent from zero (P < 0.01). Between

reporting units, FST values ranged from 0.005 to 0.411 with a mean value of 0.188. The

least diﬀerentiated populations were the fall-run populations from California’s Central

Valley, as has been observed with other genetic datasets (Williamson and May 2005;

Seeb et al. 2007).

Genetic structuring of the Chinook salmon populations in the baseline is dis-

played in an unrooted neighbor-joining dendrogram (Figure 2.1). Relationships are in

strong agreement with expectations based on geography and previous studies (Waples

et al. 2004, Beacham et al. 2006, Templin et al. 2011, Moran et al. 2013); populations

are generally organized north to south along the main branch, with populations from

within the same drainage usually clustering together. Populations from California’s

Central Valley are monophyletic relative to the remainder of the populations but are

72 characterized by short branch lengths, small distances between nodes and low bootstrap support. Central Valley spring-run and fall-run populations are also monophyletic, with the exception of the Feather River Hatchery spring-run, which is included in the fall-run reporting unit due to a history of substantial introgression between the runs and the consequent diﬃculty of genetically distinguishing them from fall-run ﬁsh (Garza et al.

2008). Sacramento River winter-run ﬁsh are quite distinct due to a well-documented recent bottleneck (Hedrick et al. 1995), and have one of the longest branches on the tree, with bootstrap support of 100%. Rivers from Northern California and coastal Oregon also form a monophyletic group. Columbia River populations are dispersed throughout the tree, although populations from the same reporting unit generally share a common branch, as do populations from Alaska.

73 Klamath_TrinityHsp

Sac_win 100 Klamath_IronGH Rogue_Applgt

Rogue_ColeRHsp CACoast_Russian 95 CACoast_Eel Snake_MCallHsumsp Yukon_Kantishna nCal_sOR_Smith 98 nCal_sOR_Chetco 91 Snake_RapidRHsumsp nPuget_KendlHsp KarlukAK COup_WellsHsumfl 93 mOR_Umpqua COmup_Wenatchee COup_Hanford mOR_Siuslaw COup_PriestHsumfl nOR_Siletz 86 nPuget_MrblHsp 100 wAK_George Snake_LyonsFHfl sPuget_SoosCH nOR_Alsea Deschutes_fl nOR_NestuccaH SitukAK 100 eVancI_BigQual COmup_CleEHsp 91 92 92 nOR_Nehalem 92 mOR_Coquille 77 98 100 COlow_CowHfl WACoast_Quinalt 75 WACoast_ForksCH 70 86 wAK_Togiak 74 COlow_CowHsp 93 CVfl_Mklmne CVfl_Deer wAK_Kanektok CVfl_UpSac_late Fraser_Harris lSkeena_Kalum to Sac_win sSEAK_Unuk COlow_KalamaHsp wVancI_RobHfl 83 CVfl_Butte mSkeena_Kitwanga SusitnaAK_Montana 57 CVfl_FeatherRHfl 100 CVfl_Mill 80 COmid_SpringCH 71 AlsekAK_Goat CVfl_Battle CVfl_FeatherRHsp Wilamette_McKenzHsp Wilamette_NSantiamH mSkeena_Morice 100 77 Taku_LilTats CVsp_Deer

CVsp_Mill 77

Thompson_SpiusCH nSEAK_PullenCH Fraser_BirkenH CVsp_UpSac CVsp_Butte 0.01

Figure 2.1: Unrooted neighbor-joining tree based on chord distances of 67 Chinook salmon populations from California to

Alaska in the GSI baseline (see Table 2.1 for population details). Dashed lines indicate the position of populations which

fall at tree junctions or have very short branch lengths. Sinona Creek and the coho salmon were omitted for missing data. 2.5.2 Assignment and Mixture Estimation Accuracy

The 7,669 individuals remaining after removal of training set fish were sub- jected to self-assignment using GSI SIM (Table 2.1). Correct assignment to population ranged from 13% for Butte Creek fall-run to 100% for five different populations.

The reporting units with the lowest correct assignment rates to population were the

Central Valley fall-run, Upper Columbia River summer-/fall-run and Lower Kuskok- wim/Western AK, averaging 28%, 36% and 40% respectively. The lowest rate of correct assignment to reporting unit was for the Siuslaw River population from the Mid

Oregon Coast, with over half of the individuals assigning to populations in the North

Oregon Coast reporting unit. The largest change in correct assignment percentage from population to reporting unit was for the Central Valley fall-run, which increased to 91%.

The results of the mixture simulations for the eight reporting units most frequently found in California and Oregon fisheries appear in Figure 2.2. Results for the remaining reporting units are not shown, as they are relatively uninformative, due to the rarity with which populations from north of the Columbia River are encountered at the southern end of the California Current marine ecosystem—an observation corrobo- rated by historical CWT data: in the three decades since 1983, only 0.5% of all CWTs recovered from Chinook salmon in California ocean fisheries were from stocks outside of California or Oregon (data from www.rmpc.org). Accurate estimates of the mixing proportions were obtained for fishery samples simulated either by CV-GC or by K-Fold.

The mean maximum likelihood estimate of the proportion of each reporting unit was

75 generally highly correlated with the true proportion, indicating that any bias was very small. For ﬁve reporting units (Central Valley fall-run, Sacramento River winter-run,

Klamath River, California Coast, and Rogue River), the 5% and 95% quantiles for reporting-unit mixing proportions corresponded closely to the quantiles one would obtain with perfect identiﬁcation of all ﬁsh (gray regions in Figure 2.2). The somewhat wider GSI quantile intervals observed for the Central Valley spring-run reporting unit were likely due to its similarity to the Central Valley fall-run reporting unit, combined with the fact that the spring-run is typically at much lower abundance than the fall-run.

Likewise, the genetic similarity of Mid Oregon Coast and Northern California/Southern

Oregon Coast fish made it difficult to accurately estimate mixing proportions for these reporting units; however, the estimates were still quite good and largely unbiased. Thus, despite the enlarged quantile intervals for Central Valley spring-run and the Mid Oregon versus Northern California reporting units, the results from both simulation methods indicated that the SNP baseline is capable of providing estimates of the true mixing proportions for most reporting units that are nearly as accurate as one would expect given perfect identification of each fish.

2.5.3 Fishery Sample

Of the 2,090 samples from California fisheries in 2010, 85 were excluded because they did not yield acceptable genotypes (< 60 successfully genotyped loci), and two samples were duplicates. Eight fish exceeded the iHz threshold of 0.56 and were removed due to potential contamination. Seven fish were identified as coho salmon

76 CentralValleysp CentralValleyfa CentralValleywi CaliforniaCoast 1.0 0.30

0.20 0.8 0.06 0.25

0.15 0.20 0.6 0.04 0.15 0.10 0.4 0.10

Estimated Proportion 0.02 0.05 0.2 0.05

0.00 0.0 0.00 0.00

0.00 0.05 0.10 0.15 0.0 0.2 0.4 0.6 0.8 0.00 0.01 0.02 0.03 0.04 0.00 0.05 0.10 0.15 0.20 0.25

KlamathR NCaliforniaSOregonCoast RogueR MidOregonCoast 0.10 0.25 0.4 0.30

0.20 0.25 0.08 0.3 0.20 0.15 0.06

0.2 0.15 0.10 0.04 0.10 Estimated Proportion 0.1 0.05 0.02 0.05

0.0 0.00 0.00 0.00

0.0 0.1 0.2 0.3 0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.01 0.02 0.03 0.04 0.05 0.06

True Proportion True Proportion True Proportion True Proportion

Figure 2.2: Estimates of mixing proportions from cross-validation over gene copies (CV-

GC) and K-Fold simulations for the eight most abundant reporting units in California

Chinook salmon fisheries. The x-axis gives the true proportion of fish from each reporting unit, and the y-axis gives the estimated proportion. The dashed line is the y=x line. Shaded regions give the range between the 5% and 95% quantiles of estimates that would be achieved with perfect assignment of fish to reporting unit; i.e., they represent the uncertainty due to the fact that fishery proportions are estimated with a finite sample (in our simulations, a sample of 200 fish). The 5% and 95% quantiles of the estimates using genetic data from the CV-GC and the K-Fold methods are shown with vertical line segments and open diamonds, respectively. The mean over 20,000 CV-GC simulation replicates and 1,000 K-Fold replicates are given by filled circles and open triangles, respectively. These points fall along the dotted line when the estimator is unbiased.

77 Table 2.3: Genetic stock identiﬁcation (GSI) results of assigning 2010 California Chi- nook salmon ﬁshery samples to their source populations using the single nucleotide polymorphism baseline, and concordance with coded-wire tag (CWT) recoveries.

# from # with # GSI/CWT % GSI/CWT Stock GSI CWT matches agreement California Coast 30 1 0 0.00% Central Valley fall 1581 958 957 99.90% Central Valley spring 7 1 0 0.00% Klamath River 108 50 49 98.00% Lower Columbia River spring/fall 1 0 0 - Mid Columbia River tule 7 2 2 100.00% Mid Oregon Coast 14 1 0 0.00% N. California S. Oregon Coast 58 25 25 100.00% Rogue River 154 11 5 45.45% Snake River fall 1 1 1 100.00% Up. Columbia River summer/fall 8 2 2 100.00% Total 1969 1052 1041 98.95% through both GSI assignment and with the species-diagnostic assay. Another 18 samples did not meet assignment conﬁdence criteria (mean z-score of -3.99 and a mean of

75 successfully genotyped loci) and were also excluded. For the remaining 1,969 fish, assignment probabilities to reporting unit ranged from 36.4% to 100% (mean 98.5%) and z-scores ranged from -4.12 to 2.68 (mean -0.04). Central Valley fall-run fish dominated the stock composition, accounting for over 80% of sampled fish, followed by the

Rogue River (7.79%), the Klamath River (5.46%) and eight other stocks with less than

5% (Table 3). Of the assigned ﬁsh, 1,052 contained coded wire tags that were recovered. Genetic assignment to reporting unit disagreed with CWT origin for only 11 ﬁsh

(1.05%) and, of these mismatches, six were ﬁsh with Klamath or Smith River tags that were assigned to the genetically similar Rogue River reporting unit.

78 2.6 Discussion

Here we describe one of the first large-scale SNP baselines for genetic stock identification of Chinook salmon and the first designed for use with fisheries in the Cal- ifornia Current Large Marine Ecosystem off the West Coast of the coterminous United

States. Chinook salmon are an economically and ecologically important species and are a major component of North Paciﬁc Ocean ﬁsheries. We genotyped over 8,000 individual

fish from 69 distinct populations at 96 SNP loci to construct the baseline. The reporting units included in the baseline represent the likely sources for over 99% of the fish typically encountered in PFMC fisheries off California and Oregon. Furthermore, mixture analyses and self-assignment indicate that the baseline has near maximum possible power for discriminating Chinook salmon stocks at the reporting unit level. Mixture proportion estimates of Central Valley fall-run, Central Valley winter-run, California

Coastal, Klamath River, and Rogue River reporting units (Figure 2.2) are no more variable than estimates that would be obtained if every fish carried an unambiguous reporting-unit tag. Mixing proportion estimates for Central Valley spring-run, North- ern California/Southern Oregon, and Mid-Oregon Coast reporting units are somewhat more variable, but still appear to be nearly unbiased. In the ocean fishery sample, assignments of over 1,000 individuals to reporting unit, using our baseline, were highly concordant (98.95%) with the CWTs recovered from the same fish. This SNP baseline therefore represents an important addition to the technologies available to Chinook salmon managers and researchers.

79 2.6.1 Methodological Considerations

Management of Pacific Ocean salmon fisheries off North America can be roughly divided into three regions: California and Oregon fisheries are managed by the Pa- cific Fishery Management Council (PFMC); fisheries in Washington, British Columbia,

Canada and southeast Alaska are subject to the international Pacific Salmon Treaty, reported to and regulated by the Pacific Salmon Commission (PSC); and fisheries further north and west in Alaska are managed by the state, with salmon by-catch under the purview of the North Pacific Fishery Management Council. The genetic baseline described here was designed primarily to identify fish caught in PFMC ocean fisheries and in ecological investigations in the southern portion of the California Current ecosystem and its associated tributary rivers and streams. We have shown that it performs well in this area but, due to an ascertainment strategy during SNP discovery that included individuals from the Columbia River and British Columbia (Clemento et al. 2011), the baseline also has sufficient statistical power to identify the source of some fish from elsewhere in the species’ North American range. We observed high rates of self-assignment to reporting unit for all regions represented in the baseline, even though some reporting units are clearly composed of populations with minimal differentiation from each other.

Moreover, the utility of our baseline could be eﬀectively extended by simply genotyping the same panel of SNPs on additional populations in those regions, even though heterozygosity and mean number of alleles (Table 2.1), and presumably statistical power, in our baseline is reduced for populations from Canada and Alaska.

80 Other SNP baselines for Chinook salmon have also been described or are being constructed. Templin et al. (2011) describe a 45 SNP locus baseline of populations in the northern and western parts of the species range, designed primarily for GSI of populations from western and southcentral Alaska. This same baseline was also used to probe the seasonal distribution and migration pattern of Chinook salmon in the Bering

Sea and North Paciﬁc Ocean (Larson et al. 2012). Despite the presence of 14 populations from California, Oregon and Washington in that baseline, the authors appropriately emphasize that resolution of those southern populations is suﬃcient only for broad-scale assignments. Similarly, Warheit et al. (2013) describe the marker selection for eventual

development of a SNP baseline for application to PSC ﬁsheries. While the existence of

multiple regional baselines is likely to expand, it will still beneﬁt the entire community of

ﬁshery managers and scientists to carefully design marker panels with as much overlap as

possible. It is conceivable that two or three panels of 96 SNPs could provide the level of

resolution needed for identiﬁcation throughout the species range. Alternatively, as next

generation sequencing techniques mature, genotyping-by-sequencing (GBS) approaches

might yield data for GSI at lower cost than current genotyping techniques. Such a

GBS approach could be used to simultaneously genotype all of the SNPs in each of the

regional baselines, allowing mixed-stock analysis throughout the species’ range.

Inclusion of the species-diagnostic marker and coho salmon sample in the base-

line provided insight into the prevalence of misidentiﬁcation of coho salmon in ocean

fisheries. In the 2010 fishery off California, seven fish sampled as Chinook salmon were

found to be coho salmon. Without such methods to identify coho salmon, they are as-

81 signed by the baseline with erroneously high conﬁdence to a northern, low-heterozygosity

Chinook salmon population (data not shown). This problem is characteristic of most statistical methods for performing GSI: if an individuals true population of origin is not included in the baseline, then even if all the populations in the baseline are very poor candidates for the fishs origin, that fish might still be assigned with high posterior probability to one of the populations. This occurs when one population is much more likely than any of the other incorrect populations, even though it is not a likely origin for the individual on an absolute scale. We introduced a simulation-based z-score method, implemented in GSI SIM, to identify fish that have likely not originated from populations in the baseline. An alternative, Bayesian nonparametric approach to dealing with

fish from populations not in the baseline identifies those fish and estimates the allele frequencies in their (unrepresented) source population (Pella and Masuda 2000). That approach is particularly appropriate when large numbers of fish are sampled from each of the populations that are not included in the baseline and when the unrepresented populations are quite divergent from all those in the baseline. We chose the z-score approach over the Bayesian nonparametric approach for three main reasons: 1) it is computationally fast and simple, as there are no convergence problems that might be difficult to detect; 2) our baseline was sufficiently comprehensive for stocks contributing to PFMC fisheries that it is unlikely that large numbers of fish would originate from any single unrepresented population, let alone a highly divergent one; 3) our approach should be more appropriate for identifying fish whose genotypes are aberrant due to genotyping complications or sample contamination. Regardless of which method is used, all GSI

82 estimation should include some analysis to identify ﬁsh that either are from populations not included in the baseline or that have aberrant genotypes for another reason.

GSI is highly dependent on source populations being sufficiently genetically differentiated from one another for discrimination. In situations where hatchery broodstock transfers, supplementation, or other processes increase straying and gene flow between fish populations, differentiation decreases and it can become more difficult to use GSI. Such is the case in the Central Valley of California, where average FST between populations in the fall-run reporting unit is 0.006 and in the spring-run reporting unit is

0.013. In the dendrogram (Figure 2.1), this region is characterized by extremely short branch lengths, small inter-nodal differences and weak bootstrap support. Extensive straying of hatchery salmon due to off-site juvenile releases (California Hatchery Scien- tific Review Group 2012) and water operations (Fisher 1994) have eliminated historical differentiation between populations of fall-run Chinook salmon (Williamson and May

2005). Introgression between fall-run and spring-run ﬁsh at the Feather River Hatchery, and likely elsewhere within the basin, has reduced diﬀerentiation between these two phe- notypes, with mean FST of 0.025 between fall-run and naturally spawning spring-run populations.

Sampling of diﬀerent stocks for baseline construction in the presence of high stray rates is not entirely straightforward, particularly when populations are largely sympatric and not visually distinguishable. For example, there is clearly a single Central

Valley fall-run ﬁsh sampled as winter-run in our baseline. These types of occurrences are almost inevitable given the high degree of disturbance and hatchery supplementation

83 over much of the species range. One approach is to move fish with discrepant genotypes from the baseline populations in which they were sampled to the ones to which they assign using GSI (e.g. Banks et al. 2000). However, such a procedure can introduce an upward bias in the predicted accuracy of the baseline, if, in fact, the removed fish actually do belong to the populations from which they were sampled, but simply have unlikely genotypes at the genetic markers used for baseline construction. We chose to be conservative in both 1) accepting a slightly lower rate of predicted resolution obtained by not removing mis-categorized fish, and 2) avoiding an upwardly biased predicted GSI accuracy if the fish removed are not mis-categorized.

2.6.2 Implications for Management

Accurately estimating the proportion of fish from different populations in mixed-stock ocean fisheries has important applications for harvest management and conservation. Stocks comingled in ocean fisheries can vary widely in productivity and abundance. Without precise information on their ocean distribution (as can be provided by GSI), managers have few options for protecting depressed or at-risk stocks from fishery impacts other than shutting down or curtailing fisheries over broad areas, as is currently done. For example, in 2008 and 2009, the largest closures on record of

ﬁsheries in California and Oregon were enacted to protect the severely reduced Central

Valley fall-run stock (Lindley et al. 2009). The economic impact of ﬁshery closures is substantial, resulting in millions of dollars of lost income for ﬁshermen, coastal communities and retailers (Michael 2010).

84 Management of Chinook salmon in California, Oregon, and Washington, and in PSC-managed ﬁsheries depends heavily on information generated by an elaborate

CWT program (Hankin et al. 2005). Tiny wire tags are mechanically implanted into the heads of juvenile fish, with each tag bearing a code that identifies the release group and source hatchery (or stock) of the fish. Tagging of naturally spawned juvenile fish has generally proven unsuccessful (Beacham et al. 1996), so tagged hatchery stocks are used as proxies to estimate fishery impacts for groups of natural stocks. Aside from the largely unvalidated assumption that such proxies accurately reflect fishery impacts on the associated natural stocks (Hankin et al. 2005), the physical effects of tagging fish and removing their nerve-rich adipose fin (Buckland-Nicks et al. 2012) as an associated

external mark can increase disease transmission (Elliott and Pascho 2001), interfere with

homing (Morrison and Zajac 1987, Habicht et al. 1998) and swimming ability (Reimchen and Temple 2004) and may impact size-at-return for adult salmon (Vander Haegen et

al. 2005). Moreover, extremely low recovery rates mean that CWT data are often quite

limited and there is frequently great uncertainty associated with the resulting estimates

derived from them (Hankin et al. 2005).

GSI has been advanced as an alternative to CWTs in ﬁshery management for

several decades. Our direct comparison of CWT and genetic assignments demonstrates

that our baseline is capable of identifying ﬁsh to reporting unit with accuracy compa-

rable to CWTs. Furthermore, using GSI, considerably more ﬁsh can be identiﬁed to

reporting unit, including ﬁsh from natural stocks. Conﬁdent genetic assignments were

obtained for ∼94% of fish from the 2010 fishery sample, whereas only 1,052 of those fish

85 carried coded wire tags and this number is inﬂated partially due to oversampling of ﬁsh believed to carry CWTs.

Fishery management decisions rely heavily upon cohort-based ocean harvest models (cf., O’Farrell et al. 2012), which require information on both stock of origin and age of fish impacted in fisheries. Since GSI does not provide fish age, it is not by itself an adequate alternative to CWTs. Nonetheless, new statistical methods capable of integrating GSI, length data, and scale- or otolith-based age data have recently been developed and shown to provide important inference in PFMC fisheries that are not available from CWTs alone (Satterthwaite et al. 2013). Moreover, pedigree-based

genetic tagging does supply age for salmon (Anderson and Garza 2006, Garza and An-

derson 2007). This method, termed parentage-based tagging (PBT), can identify the

actual parents of a genotyped individual through parentage analysis if they have been

genotyped with the same genetic markers. If the parents date of spawning is known, as

it typically is in a hatchery, then the reconstructed pedigrees yield the oﬀsprings precise

age and any associated parental spawning information.

Importantly, both PBT and GSI can be undertaken with the same SNP geno-

types, and the SNPs used in our GSI baseline are suﬃciently powerful for PBT with

salmon from California to Washington (Anderson 2012). This interoperability of geno-

type data enables an integrated program that uses both GSI and PBT simultaneously,

providing identification for all fish in a fishery or ecological sample and yielding signifi-

cantly greater inference than either method alone. For example, GSI cannot distinguish

between spring-run and fall-run ﬁsh from the Feather River Hatchery in California, but

86 PBT discriminates them, almost without error, from any mixture. Likewise, though it is difficult to implement PBT in natural populations, the same SNP genotypes used in a PBT analysis permit accurate identification (via GSI) of fish from the naturally spawning, ESA-listed California Coastal Chinook Salmon ESU.

2.7 Conclusions

The advent of high-throughput SNP genotyping has already revolutionized human genetics (Jenkins and Gibson 2001), providing previously unattainable resolution

(e.g, Novembre et al. 2008), and is poised to do the same for fisheries biology and management. Here, we use a careful and statistically valid power analysis of SNP genotypes from a large number of Chinook salmon populations concentrated at the southern end of their native range to show that SNPs can provide a powerful baseline for genetic stock identification (see also Larson et al. 2012) in fisheries and ecological investigation in the California Current and its tributaries in California and Oregon. We predict that these advances in genetic resources and methods will foster fundamental improvements in the way salmon populations are studied, monitored and managed.

87 Chapter 3

Large-scale genetic tagging experiment in a hatchery population of Chinook salmon (Oncorhynchus tshawytscha)

allows for pedigree-based inference

3.1 Introduction

Studies on natural selection, behavioral ecology and the population biology of

plants and animals often require tracking individuals, groups or populations over a pe-

riod of time. This is generally achieved by marking or tagging individuals for subsequent

recapture or detection. Physical tags have been used to elucidate the migration and dis-

persal patterns of birds over the last century (Baldwin 1921, Nickell 1968, Greenwood

and Harvey 1982, Moore and Dolbeer 1989), however, mark-recapture experiments are

88 also common in studies of ﬁsh (Metcalfe and Arnold 1997, Jones et al. 1999), mammals

(Hoskinson and Mech 1976, Ormiston 1985, Bethke et al. 1996) and even insects (Stern et al. 1965, Sumner et al. 2007). These tagging experiments have been used for a broad range of applications, which include: investigating behavioral responses to changing conditions, estimating the eﬀects of natural selection and delineating the distribution of populations.

While physical tagging has a long history, the increasing availability of genomic resources has made genetic tagging methods (Palsbøll 1999) a viable alternative for a variety of species. In its simplest form, genetic tagging is analogous to physical tagging, with the ‘mark’ being the ﬁrst time a genotype is encountered and a ‘recapture’ occurring by matching the original genotype to a subsequent sample. The use of genetic information in lieu of traditional tags to identify individuals has been demonstrated in taxa as diverse as whales (Palsbøll 1997), bears (Woods et al. 1999), and martens

(Mowat and Paetkau, 2002). Genetic information can also be used to assign membership of individuals to their most likely population of origin. This method, termed genetic stock identification (GSI) in fisheries, and called population assignment in the field of molecular ecology, requires collecting baseline allele-frequency data from potential source populations and then uses maximum likelihood (Smouse et al. 1990) or Bayesian methods (Pella and Masuda 2000) to determine the probability that the sample originated from each population; sample and baseline genotypes are collected for the same set of genetic markers. GSI has been successfully applied in studies of highly-structured salmon populations for almost three decades, but is limited if groups are not sufficiently

89 diﬀerentiated (Beacham et al. 1985, Teel et al. 1999, Beacham et al. 2006, Seeb et al.

2011, Clemento et al. in review).

Yet another way that genetic data can be used as a tagging methodology is in the inference of relationships between individuals, primarily first order relatives such as parent-offspring or siblings. Parentage analysis has been used to address a diverse range of ecological questions, including dispersal, hybridization, fitness, relatedness and estimation of population size (DeWoody 2005). In an early genetic mark-recapture experiment in turtles, parentage was used to reconstruct and subsequently recapture a paternal genotype that was not directly observed (Pearse 2001). Many methods are also available to reconstruct sibships between individuals without parental information

(e.g. Wang 2004), as well as to identify parents and oﬀspring in the wild (Jones et al.

2009). Assignment of parentage with molecular markers generally utilizes Mendelian incompatibilities between oﬀspring and putative parents to exclude unlikely trios, since a true oﬀspring must carry one of the maternal and one of the paternal alleles. There is a variety of software available for actually using genetic data to infer parentage, however, many are limited in their computational capacity and ability to handle large and complex datasets in a reasonable amount of time (Jones et al. 2009). However, new algorithms

have now been developed to perform truly large-scale parentage inference, allowing for

practical extension of these genetic tagging methods to high fecundity organisms like

salmon (Anderson and Garza, 2006, Anderson, 2012). Additionally, recent development

of large numbers of single nucleotide polymorphism (SNP) markers (Clemento et al.

2011, Abad´ıa-Cardoso et al. 2011), which are amenable to eﬃcient high-throughput

90 genotyping, now allow for the practical analysis of the large number of individuals in salmonid populations (Abad´ıa-Cardoso et al. 2013, Steele et al. 2013).

Chinook salmon (Oncorhynchus tshawytscha) are a highly valued species of

Pacific salmonid and are the target of large commercial and recreational fisheries. Chi- nook salmon are anadromous, wherein adult fish migrate from the ocean to spawn in their natal river, and must therefore contend with impacts in both freshwater and marine environments. Over the last century, many Chinook salmon populations have been reduced or even extirpated by the construction of large dams, extensive water extraction for agriculture and human consumption, overfishing and variable ocean conditions

(Myers et al. 1998). This has resulted in listings under the Endangered Species Act

(ESA; FedReg 1990, 1999), particularly in the southern portion of the species range

(e.g. California, Oregon and Washington). In order to mitigate for the multiple impacts threatening Chinook salmon populations, state and federal agencies now produce millions of fish annually in hatcheries. These hatchery fish – primarily intended to reduce variability in ocean abundance and provide fishing opportunities – comingle with wild fish in the ocean and can compose the majority of the catch in certain times and places.

Ensuring sustainability and the persistence of salmon populations while providing fishing opportunities can be a complex task. Overestimation of the contribution from specific stocks can have serious conservation implications, while underestimation can leave the resource underexploited, both potentially costing the fishing industry and coastal communities millions of dollars (Michael, 2010). Generally, management of Pa-

91 cific salmon ocean harvest in the coterminous United States falls under the purview of the Pacific Salmon Commission (PSC) and Pacific Fishery Management Council

(PFMC), while NOAA Fisheries is responsible for controlling harvest of threatened and endangered population segments under the ESA. These entities employ a variety of methods to set fishing areas and seasons, determine quotas and legal gear and establish catch limits and size restrictions. Stock-specific forecasting models are used to estimate ocean abundance indices, which are then used to set harvest limits, first at the international level (PSC) then for local ocean fisheries and finally for terminal fisheries in rivers (Hyun, 2012). The accuracy of these cohort-based models and the resulting abundance forecasts are highly dependent on the quantity and quality of data; estimates of age and stock specific mortality rates and their distribution in the fishery catch, are critical inputs to these models. Currently, the primary source of information for fishery management comes from coded wire tagging of a limited number of hatchery stocks.

The need to identify stock-specific fishery impacts led, in the 1950s, to clipping of particular fins (adipose, anal, maxillary), in an attempt to identify production from different hatcheries or regions. By the 1970s, managing agencies began to use cohort information in fishery management models and turned to the use of coded wire tags

(CWTs) in juvenile ﬁsh to indicate stock and cohort of origin (Jeﬀerts et al. 1963).

CWT data has been used to estimate “exploitation rates by age, maturation rates,

adult equivalents, marine survival rates, total mortality” and even to infer exploitation

patterns of untagged natural stocks (Morishima 2004). CWTs are small pieces of metal

(0.5 - 1mm long) mechanically implanted into the heads of juvenile ﬁsh. Each tag bears a

92 group-specific code that identifies the release cohort and source hatchery (or stock). Tag recovery is accomplished through identification of fish carrying a tag (usually removal of the adipose fin), followed by removal of its head and shipment to a laboratory, where the tag is manually extracted and read under a microscope. “Harvest from a cohort is [then] estimated by expanding the number of CWTs recovered according to the fraction of the catch sampled, the fraction of the cohort carrying CWTs, the fraction of heads from recaptured fish that reach a laboratory, and the fraction of dissected heads from which a CWT is decoded (Bernard and Clark 1996).” However, due to limited tag recoveries

(often less than 1%) and assumptions about the equivalence of tagged and untagged

ﬁsh, there is frequently great uncertainty associated with the output of management models (Hankin et al. 2005).

Prior to 1996, only fish with CWTs were given adipose fin clips, but another major challenge to the continuing use of CWTs are recent state and federal regulations, which require adipose fin clips on a majority of hatchery production (Hankin et al.

2005). This will increasingly result in large numbers of adipose fin-clipped but untagged salmon and has already “decreased the effectiveness of the current program, added costs without gaining information, increased the numbers of fish that samplers handle and mutilate[,] and decreased the value [of] these fish to retailers (Alexandersdottir et al.

2004).” This problem has necessitated the use of secondary, electronic tag detection methods at considerable increased cost and effort to the entire program. CWTs are also subject to loss at uncertain rates, which effectively increases the number of clipped but untagged fish (Johnson 2004). Moreover, the tagging and marking process may

93 cause subtle injuries to juvenile ﬁsh that can aﬀect performance and survival at later life stages (Morrison and Zajac 1987, Habicht et al. 1998, Reimchen and Temple 2004).

Given the declining eﬀectiveness of the current CWT program, the PSC has recommended validation of alternative tagging strategies (Hankin et al. 2005). One of

the most promising technologies, and the one evaluated here, is parentage-based tagging

(PBT; Garza and Anderson 2007). Utilizing a novel statistical genetic framework for

large-scale parentage analysis, genotypes collected from parental breeding generations

in hatcheries are used to tag the oﬀspring cohort. Subsequent non-lethal sampling of

ﬁsh during their seaward migration, in ﬁsheries, or upon return to spawn (either at

hatcheries or instream) is followed by high-conﬁdence parentage assignment (Anderson

2012), allowing accurate pedigree reconstruction, and identifying stock and cohort of

origin in the process. Since a pair of Chinook salmon can produce thousands of oﬀ-

spring, the tagging of juveniles through genotyping of parents is highly eﬃcient. This

methodology generates the same information as the current coded wire tag (CWT) pro-

gram, which currently provides the bulk of the cohort-speciﬁc ﬁshery mortality data for

salmon in the northeast Paciﬁc. The ability to accurately identify oﬀspring of spawn-

ing ﬁsh through parentage analyses means that a pair of parental genotypes translates

into many genetic tags in the next generation and has broad potential application for

population assessment of ﬁsh and other high fecundity species.

Described here is a large-scale, intergenerational genetic tagging experiment

with a hatchery population of Chinook salmon from the Feather River, CA, USA. I

ﬁrst examine whether the same panel of SNP markers, successfully used to construct

94 a coastwide baseline for GSI (Chapter 2; Clemento et al. in review), can also be used to confidently reconstruct pedigrees of individuals that have undertaken an ocean migration. The accuracy of assignments is determined by comparing them with recorded cross information in order to evaluate whether genetic tagging data is comparable with that derived from the physical tags currently deployed in the system. Reconstructed parent-offspring trios are used to assess interannual variability in the age structure of offspring cohorts as well as the age structure and relative reproductive success (i.e variation in family size) of spawning broodstock. Data on the physical characteristics of parents and offspring allow for estimates of the heritability of length at maturity and correlations between female body size and the number of her offspring returning to spawn. Inbreeding and relatedness in spawning populations is assessed and the effects of parental relatedness on their reproductive success is evaluated. This research also provides the first evidence that PBT can identify parentage of offspring in large mixed-

fishery samples. I demonstrate that parentage-based genetic tagging provides not only a powerful and efficient means of tagging large numbers of individuals, but also generates novel population information that can be used to inform hatchery and fishery management.

95 3.2 Methods

3.2.1 Study Site

The Feather River is one of the largest tributaries to the Sacramento River in the northern part of California’s Central Valley. Historically, the Feather River supported runs of both the fall-run and spring-run ecotypes of Chinook salmon. The spring-run phenotype is characterized by adults that are sexually-immature when they migrate upstream during the Spring. These ﬁsh hold in deep pools throughout the summer and then mature and spawn and die throughout the fall and early winter months.

In the fall-run phenotype, sexual maturation is coincident with upstream migration and spawning during the fall months. Prior to human modification of the watershed, spring- run fish spawned in the upstream reaches of the Feather River, spatially separated from the fall-run fish spawning further downstream (Department of Water Resources 2004).

In 1968, however, construction of Oroville Dam, a principal feature of the California

State Water Project, was completed on the mainstem Feather River. This dam blocks upstream passage of spring-run Chinook salmon (Fry and Petrovich 1970), conﬁning them to spawn in the same downstream reaches where fall-run Chinook also spawn.

As a consequence, introgression between the two types has been widespread on the currently available spawning grounds (Yoshiyama 1996, Williamson and May 2005).

Additionally, propagation of Chinook salmon at the Feather River hatchery has contributed to introgression between the two run types. For the ﬁrst three decades of operation at the hatchery, little was done to distinguish or isolate the two run types:

96 mature ﬁsh were simply spawned as they arrived at the hatchery. Fish that entered the hatchery in September were considered to be “spring-run” and were spawned together, while those that entered in October were spawned together as “fall-run” (Department of

Water Resources 2004). This practice did little to maintain the reproductive isolation of the two runs because both spring run and fall run ﬁsh typically mature between the months of October and December. Since 2003, the California Department of Fish and Wildlife (CDFW), who operates the Feather River Hatchery (FRH; Oroville, CA), has made a concerted eﬀort to limit the amount of introgression between the two runs.

Specifically, they devised a plan that is meant to exclude potentially fall-run fish from breeding with fish displaying the spring-run phenotype. During May and June at the hatchery, early-arriving, sexually immature fish are marked with an externally visible tag and released back into the river. Fish that arrive after July 1 are not admitted to the hatchery and so remain in the river untagged and are ultimately assumed to be fall-run. Fish are again allowed to swim up the ladder into the hatchery in late

September/early October where they are sorted based on the presence or absence of the external tag which identiﬁes individuals that expressed the spring-run phenotype.

Tagged, early-arriving females are then mated one-to-one with spring-run males, the eggs incubated in daily lots, and the ﬁsh subsequently reared to the fry life stage and released in various locations in the drainage.

97 3.2.2 Hatchery Sampling

Caudal ﬁn clips were collected from all returning ﬁsh (spawned and unspawned) by CDFW personnel at the FRH and dried on blotter paper. Comprehensive sampling and genotyping of the spring-run Chinook broodstock took place for the six years from

2006 to 2011 (Spring-run/Spring-origin in Table 3.1), while the fall-run broodstock (Fall- run/Non-spring-origin in Table 3.1) was also genotyped in 2008. Coded-wire tag data was used to retrieve samples from spring-run offspring that were collected as fall-run spawners in 2009, 2010 and 2011 (Fall-run/Spring-origin in Table 3.1), for assignment to spring-run parents. A small subset of 2012 spring-run fish, whose offspring were to be used to reintroduce Chinook salmon to the San Joaquin River, were also analyzed for parentage. Metadata, including gender, spawn date, fork length (mm) and spawning partner (spring-run only, 2006-2009) was recorded for each fish. In total, samples from 12,817 Feather River Hatchery Chinook salmon were collected and genotyped (Ta- ble 3.1).

98 Table 3.1: Summary of sampling and genotyping effort and success at the Feather River Hatchery, Oroville, CA. Included are the year of spawning, the hatchery designation of the run (Spring or Fall), the number of individuals genotyped (males and females), the number of individuals excluded for missing genotypes at more than 10 loci, and the number of individuals spawned, as reported by the hatchery. Population genetic parameters of unbiased heterozygosity (Hz), observed heterozygosity (Ho), the inbreeding coefficient (FIS; values with an asterisk are significantly different from zero, p<0.05, 1000 permutations) and the mean individual relatedness coefficient (Rxy; see text) were calculated for each genotyped broodstock year and spawn run. Spring origin fish among Fall-run spawners were identified using coded-wire tag data and included as putative offspring during parentage analysis.

Genotyped Excluded Matings Mean Spawn Spawn Run n n n n n n indiv. Year Run Origin [ ][ ] [ ][ ] [ ][ ] Hz Ho FIS Rxy 2006 Spring n/a 593♂ 553♀ 144♂ 47♀ 590♂ 590♀ 0.375 0.377 -0.0057 -0.0066 Fall n/a -- -- 3390 3431 ---- 99 2007 Spring n/a 731 692 82 86 701 701 0.374 0.373 0.0031 -0.0046 Fall n/a -- -- 1432 2233 ---- 2008 Spring Spring 711 718 268 200 387 390 0.370 0.368 0.0044 0.0064 Fall non-Spring 1572 1716 297 154 1463 1680 0.370 0.367 0.0082* 0.0074 Fall Spring ------2009 Spring Spring 357 464 12 3 399 480 0.372 0.369 0.0077* 0.0005 Fall non-Spring -- -- 2838 2946 ---- Fall Spring 268 359 31 17 ------2010 Spring Spring 711 615 73 50 611 611 0.373 0.374 -0.0038 -0.0004 Fall non-Spring -- -- 7628 4961 ---- Fall Spring 388 344 53 54 ------2011 Spring Spring 551 577 9 18 529 529 0.371 0.375 -0.0116 0.0031 Fall non-Spring -- -- 6141 6887 ---- Fall Spring 327 388 75 92 ------2012 Spring Spring 90 92 0 1 90 90 ---- Total 12817 1766 mean 0.372 0.372 0.0003 0.0008 3.2.3 DNA Extraction and Genotyping

Collected tissue was sub-sampled and furnished to the SWFSC Santa Cruz Lab for analysis. DNA was extracted from dried tissue using Qiagen DNEasy 96 kits on a

BioRobot 3000 (Qiagen, Inc.) according to the manufacturer’s recommended protocols.

All individuals were then genotyped at the 96 SNP loci described in Chapter 2 (Table 2.2.

A multiplex pre-ampliﬁcation reaction was used to increase copy number of targeted genomic regions. Unlabeled primers (no ﬂuorescent probes) for the panel of 96 loci were combined and diluted to 50nM; the 5uL multiplex PCR contained 1.25uL of this pooled assay mix, 1.25uL of extracted DNA and 2.5uL of 2X Multiplex Master mix

(Qiagen). The pre-amp thermal cycling routine consisted of 95◦C for 15 min followed

by fourteen cycles of 95◦C for 15 seconds and 60◦C for 4 minutes and a ﬁnal hold at 10◦C.

Multiplex PCR product was diluted with 15uL of 2mM Tris and frozen. Samples were

then genotyped on 96.96 Dynamic arrays (Fluidigm Corporation) using a Fluidigm EP1

according to manufacturer’s protocols. Genotypes were called and the data collected

using the Fluidigm SNP Genotyping Analysis software (vers. 2.1.1). Individuals with

missing data at 10 or more loci were excluded from further analyses.

3.2.4 Population Genetic Analyses

Observed (Ho) and unbiased expected heterozygosity (Hz; Nei 1987) were cal-

culated for each brood year using the Microsatellite Toolkit (Park 2001). The inbreeding

coeﬃcient (Fis), a measure of increased homozygosity due to inbreeding, was calculated

for each brood year using the software package Genetix (Belkhir 2004) and signiﬁcance

100 assessed with 1000 permutations of the dataset.

3.2.5 Pedigree Reconstruction

Upon release as juveniles (yearlings) back into the Feather River, Chinook salmon from the FRH migrate to the ocean and then return to spawn at age two, three and four. As such, the spring-run broodstock from 2006, 2007, 2008 (including fall-run),

2009 and 2010 was used as the potential parents of ﬁsh returning to spawn in 2008, 2009,

2010, 2011 and 2012. The software package SNPPIT (Anderson, 2012) was employed to perform parentage assignments. SNNPIT is a powerful and eﬃcient tool for assigning parentage, which proceeds in two steps. First, the software assembles all possible pairs of parents and uses Mendelian exclusion to exclude pairs that cannot be the parents of the individual to be assigned. Each oﬀspring is then assigned to the most likely parent pair from amongst those with few enough Mendelian incompatibilities. The software then employs Monte Carlo simulation with a novel importance sampling algorithm to calculate a p-value and associated false discovery rate (FDR) for each parentage assignment. Genotyping error rate was assumed to be 0.005 per gene-copy (1% per locus) for a majority of the loci used. Using observed Mendelian incompatibilities in reconstructed trios, however, genotyping error rates were estimated directly for four loci, adjusting the value for Ots AldB1-122 to 0.0094, Ots 105401-325 to 0.0265, Ots 112208-722 to

0.027 and Ots 101704-143 to 0.011.

Each brood year from 2008-2012 was assigned parentage separately, however, all previous years were included as potential parental sources. This provided a test for

101 false-positive assignments, as the life-history of these animals (returning as two-year olds at the youngest) should preclude assignments to the year directly preceding the brood being assigned. No parentage assignments are expected to indicate that a returning spawner is only one year old.

Two independent SNPPIT runs were conducted for each group of offspring being assigned to parents. In the first run, metadata on spawn date and sex of the parents was ignored, such that all possible pairs of individuals within a year were considered possible parent-pairs, even if it was reported that they were spawned on different days or were the same sex. The second run limited possible parent pairs to only males and females spawned on the same day. Comparison of the two runs identified some minor metadata errors, and additionally verified correct assignments. Assignments for individuals with an FDR > 0.01 were conservatively excluded, meaning that no more than one assignment in a hundred is expected to be incorrect by chance alone. Parentage assignments were compared to recorded crosses for the spring-run years 2006-2009.

3.2.6 Age Structure, Reproductive Success and Length-at-spawning

The age of returning adults was determined for the 2008-2011 spring-run broodstock and the small sample from 2012 and ranged from two to four. Offspring from fish spawned in 2006 (hereafter, the 2006 cohort) could be identified when they returned at age two, three, four and five in 2008, 2009, 2010 and 2011, respectively; fish from the 2007 cohort could be identified returning at age two, three, four and five in

2009, 2010, 2011 and 2012, respectively; ﬁsh from the 2008 cohort could be identiﬁed

102 returning at age two, three and four in 2010, 2011 and 2012 respectively; and fish from the 2009 cohort could be identified at age two in 2011 and age three in 2012. The proportion of fish returning at age two, three and four years old from the 2006 and 2007 cohorts was compared using z-tests. Note that the 2012 sample is only a small subset of the total 2012 spring broodstock, and therefore the 2008 cohort was excluded from this analysis as one would expect four year old fish to be under-represented.

Parentage assignments were also used to examine the age structure of the spring-run spawners in 2010 and 2011. This ﬁrst required removing the individuals that were included in these groups from the fall-run spawn groups as potential oﬀspring of

2006, 2007, 2008 and 2009 parents (fall-run/spring-origin). For the remaining individuals (those actually spawned as spring-run in 2010 and 2011), the number of individuals that were age two, three and four were identiﬁed and z-tests used to compare the relative proportions of the three age classes between years.

The distribution of family size was examined using the inferred parent-offspring trios for fish spawning in 2006, 2007, 2008 and 2009. This analysis included only those parents with at least one offspring detected via pedigree reconstruction. Reproductive success was estimated by counting the number of offspring per parent that returned to the hatchery in any year. As the number of offspring per parent pair was not normally distributed, a Kruskal-Wallis test was used to detect differences between the 2006 and

2007 cohorts for which all age classes (two, three and four year olds) were likely observed.

While the hatchery reports primarily one-to-one matings (only one male and one female per cross), the relative reproductive success of males versus females versus pairs was

103 examined across years, again including only those families with one or more oﬀspring.

Additionally, parentage reconstructions were used to discern the structure of sibships in the 2012 sample to be used as broodstock for reintroduction to the San Joaquin River.

Because the hatchery records the length of each fish, and parent-offspring relationships were identified by parentage analysis, the heritability of length-at-spawning was investigated for the dominant three-year old age class. The slope of the parent- offspring regression line was used to estimate heritability (h2). The mean length of

each parent pair was compared to all of their oﬀspring and to male and female oﬀspring

separately. The relative contribution of fathers’ and mothers’ lengths to oﬀspring length

was also analyzed separately, speciﬁcally looking at the contribution of fathers to sons

and mothers to daughters. Since larger females also generally produce a larger number

of eggs, the length of each mother was compared to her reproductive success and the

regression again ﬁt with a linear model.

3.2.7 Relatedness

For each collection of spring-run spawners (2006-2011) and the single collection

of fall-run spawners (2008), the relatedness coeﬃcient (Rxy of Queller and Goodnight

1989) was calculated between all pairs of individuals in each collection using the software

KINGROUP (Konovalov et al. 2004). Rxy provides a measure of the probability that the shared alleles between two individuals are identical by descent (IBD); higher values of Rxy suggest an increased degree of relatedness with a maximum value of 1 indicating identical genotypes. For each sample, a histogram of Rxy values was plotted, and the

104 mean, standard deviation and skew calculated. The distribution of Rxy values was also compared to a normal distribution with same mean and standard deviation as that observed in the sample. Since the hatchery kept records of the matings from 2006-2009, the distribution of Rxy values among pairs that achieved reproductive success (defined as those that had at least one offspring return to the hatchery in a subsequent year) could be compared to those that did not. Again, the mean, standard deviation and skew were calculated and the distribution of Rxy values for successful versus unsuccessful matings plotted. As relatedness data appeared to be normally distributed, a two-sided t-test was used to examine whether the mean of Rxy values were significantly different for successful versus unsuccessful parent pairs, for each year and over all years. Finally, the correlation between the size of each full-sib family and the degree of relatedness (value of Rxy) between the parents was investigated with a simple linear regression.

3.2.8 Fishery Samples

In 2010, CDFG collected samples at California ports from 2090 salmon landed in commercial and recreational fisheries. About half of these fish carried CWTs that identified their population of origin and age. Using SNPPIT, the FRH broodstock collections were searched for parents of these port-sampled fish and compared to the genetics-based recaptures to the CWT data. Again, assignments with FDR values >

0.01 were excluded as low confidence. Samples were also collected from the commercial salmon fleet in 2010, 2011 and 2012, primarily for analysis with a genetic stock identification (GSI) baseline (Clemento et al. in review), however these collections also contain

105 oﬀspring from the FRH hatchery. CWT data were unavailable for the GSI collections.

In total, DNA was extracted from 24,242 ﬁshery samples, genotyped with our panel of

96 SNP loci and examined for parentage among the sampled FRH brood stock samples.

The age structure of FRH ﬁsh in the four ocean samples was determined using the parentage reconstructions.

3.3 Results

A total of 12,817 Chinook salmon collected at the Feather River Hatchery from

2006 to 2012 were genotyped with our panel of 96 SNP loci (Table 3.1). Genotypes from 1766 samples were excluded due to missing data (>10 missing loci), leaving 11,051 samples for further analysis. These analyzed samples fell into three categories: spring- run spawners (sample sizes ranged from 181 in the partial 2012 sample to 1255 in 2007 with a mean of 923); fall-run spawners (2008 only, with a sample size of 2837); and fall-run spawners of spring-run origin as determined by CWT data from 2009 to 2011

(with an average sample size of 584 per year). The last category of individuals (fall- run spawn/spring origin) were included as potential oﬀspring, but were excluded from the parent broodstock sample for pedigree reconstruction and calculation of population genetic statistics.

3.3.1 Population Genetic Parameters

Estimates of unbiased heterozygosity ranged from 0.370 in the 2008 samples to 0.375 in the 2006 sample and averaged 0.372, while observed heterozygosity ranged

106 from 0.367 in the 2008 fall-run collection to 0.377 in the 2006 collection with a mean of

0.372 (Table 3.1). The inbreeding coefficient, Fis, ranged from -0.0116 in 2011 to 0.0082 in the 2008 fall-run sample; values for both the fall-run sample and the 2009 sample were significantly different from zero (P < 0.05). The overall degree of relatedness was estimated by first calculating the mean value of Rxy between each individual and all other individuals and then taking the mean of these individual values in each collection.

This mean individual relatedness ranged from -0.0066 in 2006 to 0.0074 in the 2008 fall-run sample and averaged 0.0008 over all collections.

3.3.2 Hatchery Pedigree Reconstruction

Two independent pedigree reconstructions were performed: in the first, assem- bly of the possible parent pairs was not limited to individuals of the opposite sex or with the same reported spawn date, while in the second these factors were used to limit the space of possible parent pairs. In the analysis unconstrained by gender or spawn date, a total of 2791 parent-offspring trios were identified. Fifteen of these trios were not present in the pedigree reconstruction limited by gender and spawn date; three assignments were to parents with different spawn dates while the remaining assignments identified two parents of the same gender. These assignments were of high confidence, with low FDR scores (mean, 0.0028) and high maximum posterior probabilities (mean,

0.99), indicating that they are likely correct trios with errant metadata. Additionally, for three of the unique parent pairs in this group multiple oﬀspring were assigned, further supporting the idea that they were true parental pairs, however, without additional

107 information to resolve the apparent discrepancies, these 15 assignments were excluded from further analyses.

The remaining 2776 parent-oﬀspring pairs had FDR values ranging from 0 to

0.0098 (mean, 8.81×10−5), with p-values ranging from 0 to 0.04 (mean, 5×10−4) and

posterior probabilities of the parent/oﬀspring trio relationship ranging from 0.5418 to

0.9999 (mean, 0.9934). An FDR of 0.0098 can be interpreted as an expectation of 27

misassignments (0.98% of the 2776), although only 13 assignments exceeded an FDR

of 0.002 (at which only 6 assignments are expected to be incorrect). In neither the

constrained nor the unconstrained pedigree reconstruction were any oﬀspring assigned

to parents from the immediately preceding year, suggesting a low false positive rate. The

assignment of oﬀspring to the correct parent pairs was also conﬁrmed by the hatchery

recordings of the mated individuals. Parentage assignments recovered 1203 correct

parental pairings from among the 1874 recorded at the hatchery (64.2%). Additionally,

354 of the recorded pairs not identiﬁed by parentage analysis were from 2009, for which

there was only a limited sample of 2012 individuals that would comprise the dominant

three-year old age class (Table 3.2).

108 Table 3.2: Summary of offspring (offs.) recoveries using PBT from four spawn years of Chinook salmon from the Feather River Hatchery, CA. Reported are the number (n) of males and females with recovered offspring, the proportion of parent pairs included in the parent database (see text), the number of mated pairs recorded at the hatchery and the number of those matings confirmed by parentage analysis. A scaled estimate of the number of expected recoveries had all parents been included in the database is also shown.

109 Males Females % Parent Recorded Confirmed Offs. recovered [# analyzed] in year Spawn with with Pairs Mates Mates 2008 2009 2010 2011 2012 Total Year Offs. Offs. Included [hatchery] [PBT] [3808] [1385] [1828] [1649] [181] [8851] 2006 233 234 0.660 237 155 3 503 16 0 0 522 2007 459 459 0.783 608 520 0 47 1240 213 0 1500 2008 261 261 0.665 658 510 0 0 26 562 4 592 2009 126 127 0.964 372 18 0 0 0 34 128 162 sum 1079 1081 sum 1874 1203 3 550 1282 809 132 2776 scaled 5 822 1647 1153 139 3766 The 2776 offspring assigned parentage accounted for 31.4% of the potential

8851 offspring sampled at the hatchery from 2008 to 2012. However, 3800 of the unassigned offspring were from the large, fall- and spring-run sample in 2008, for which only the small two-year old age class could be identified (as offspring of 2006 spawners).

Excluding the 2008 offspring and evaluating only the years for which the parents of putative three-year olds were available (2009-2012), parentage was assigned for 55% of offspring. It must also be considered that a substantial number of parent pairs were not available for parentage assignment because they were excluded prior to analysis for excessive missing data (>10 loci; Table 3.1). For each day of spawning in each year, the number of genotyped parent pairs (all males x all females) was calculated and sub- tracted from the number of excluded parent pairs (excluded males x excluded females), weighted by the proportion of females spawned on that day. Summed over the spawn year, this provided an estimate of the percentage of parent pairs included in the parent database for analysis. As offspring are not assigned to single parents here, each parent pair excluded for missing data was a missed opportunity to assign parentage to an offspring. As the most likely source of missed assignments, the proportion of parent pairs in the database was used to scale observed offspring recoveries (Table 3.2). For example, had all parent pairs been included in the parent database for assignment of the 2009 offspring, one could expect to have recovered parentage assignments for 822 individuals. Using the scaled estimates of offspring recoveries from 2009 to 2012, the analysis is expected to have assigned parentage for an additional 985 fish, or a total of

3761 ﬁsh, which would be 74.58% of the 2009-2012 oﬀspring available for recovery.

110 3.3.3 Age Structure

Using the reconsructed pedigrees, the age at which ﬁsh return to spawn was assessed for the 2006 and the 2007 cohorts. Of the 522 ﬁsh assigned to parents from

2006, three (0.57%) returned at age two (100% males), 503 (96.4%) at age three (36.8% males and 63.2% females), and 16 (3.07%) at age four (43.8% males and 56.2% females).

Of the 1500 fish assigned to parents from 2007, 47 (3.13%) returned at age two (97.9% males and 2.1% females, 1240 (82.67%) at age three (53.7% males and 46.3% females), and 213 (14.2%) at age four (35.2% males and 64.8% females; Figure 3.1). While z-tests identified significant differences between the two cohorts for the proportion of two-year olds (z = -3.24, P < 0.01), three-year olds (z = 7.81, P < 0.01), and four-year olds

(z = -6.91, P < 0.01), both mixtures were dominated by the three-year old age class.

Two-year old females were uncommon in both cohorts.

Again utilizing the reconstructed pedigrees, the full age structure of the spring- run spawning broodstock was examined for the years 2010 and 2011. However, the tallies of parentage assignments for these two years in Table 3.2 contain individuals that were actually spawned with the fall-run (fall-run/spring-origin from Table 3.1). In order to get a true picture of the relative proportion of two-, three-, and four-year olds in the

2010 and 2011 spawning populations, parentage assignments of fall-run/spring-origin individuals were removed and the age structure evaluated anew. After exclusion of the fall-run spawners for 2010, 814 assignments remained, representing 61.5% of the spring- run spawners in that year. For 2011, 529 assignments were retained, representing 46.9%

111 1.0 Age-two Age-three Age-four

0.5 Frequency

0.0 Females Males Females Males (327) (195) (713) (786) Cohort 2006 Cohort 2007

Figure 3.1: Age structure of returning adults (male and female) for two cohorts (2006 and 2007) from the Feather River Hatchery, CA. Numbers in parentheses indicate the total number of ﬁsh in each category, while white bars denote two-year olds, grey bars three-year olds and black bars four-year old ﬁsh.

112 of the spring-run spawners. Among the 2010 spring-run spawners, six (0.74%) were age two (100% males), 799 (98.16%) were age three (54.1% males and 45.9% females), and nine (1.10%) were age four (44.4% males and 55.6% females). Among the 2011 spring-run spawners, 16 (3.02%) were age two (100% males), 347 (65.6%) were age three (52.2% males and 47.8% females), and 166 (31.4%) were age four (33.1% males and 66.9% females; Figure 3.2). Z-tests also detected signiﬁcant diﬀerences between the two spawn groups for the proportion of two-year olds (z = -16.1129, P < 0.01), three-year olds (z = 16.493, P < 0.01), and four-year olds (z = -3.23, P < 0.01).

3.3.4 Variance in Family Size and Reproductive Success

Parentage reconstruction yielded 2776 parent-offspring trios derived from 1083 unique parent pairs and distributed in 1081 pedigrees (only two males were found to have spawned with multiple females over the study period). A total of 1079 males and 1081 females successfully produced offspring that returned to the hatchery as adults. The mean number of offspring for successful parent pairs was 2.6 (range, 1-13; Figure 3.3).

Among successful parent pairs, 37.9% had only a single oﬀspring return and only one parent pair yielded thirteen oﬀspring, the largest full-sibling family detected. For the

2006 cohort, 39.6% of the number of hatchery-reported matings (Table 3.1) yielded recovered offspring, while 65.5% of the reported number of 2007 spawners achieved reproductive success. In 2007, more parent pairs had two offspring return than one offspring, otherwise the distribution of family sizes across years was comparable (Figure 3.4). A significant difference in the pattern of reproductive success was found between the 2006

113 1.0 Age-two Age-three Age-four

0.5 Frequency

0.0 Females Males Females Males (372) (442) (277) (252) Spawners 2010 Spawners 2011

Figure 3.2: Age structure of spawning adults (male and female) for two years of spawner broodstock from the Feather River Hatchery, CA. Numbers in parentheses indicate the total number of ﬁsh in each category, while white bars denote two-year olds, grey bars three-year olds and black bars four-year old ﬁsh.

114 and 2007 cohorts (chi-squared = 44.67, P < 0.001). Among the over 2000 ﬁsh used as broodstock for a species reintroduction to the San Joaquin River, 102 pedigrees were assembled, containing a single family of four full-sibs, four families of three full-sibs, and 19 families of two full-sibs, with the remainder as singletons.

For salmon, as with most ﬁshes, larger female body size generally allows for production of a larger number of eggs (Groot and Margolis 1991). If an increase in the number of eggs provides more opportunities to have oﬀspring return, a correlation between female body size and reproductive success may be expected. Using the estimates of reproductive success from the reconstructed pedigrees and multiple regression,

I found a highly signiﬁcant correlation (P < 0.001) between reproductive success and mothers with lengths greater than 787mm (Figure 3.5).

3.3.5 Heritability of Length-at-spawning

Using the reconstructed families and the known lengths of sampled fish, the following regressions on length were examined for 3-year old offspring: parental mean- all offspring, parental mean-male offspring, parental mean-female offspring, father-son, mother-daughter (Figure 3.6). A positive, highly significant correlation was detected for all comparisons, however variability was also high. For all 3-year old offspring, the mean parental length explained approximately 3% of the observed variation (F1,2302 =

68.59, R2 = 0.029, P < 0.001). The mean length of the parent pair explained more of

2 the observed variation in the length of female oﬀspring (F1,1160 = 43.39, R = 0.036,

2 P < 0.001) than of male oﬀspring (F1,1139 = 22.8, R = 0.020, P < 0.001). Among

115 Relative reproductive success for parents with one or more offspring 0.4

Pairs

Pas

Mas 0.3 0.2 Frequency 0.1 0.0 1 2 3 4 5 6 7 8 9 10 11 12 13

Number of Offspring

Figure 3.3: Number of oﬀspring that returned to the hatchery for females (white bars), males (grey bars) and mated pairs (dark bars) over all study years. The similarity over comparisons is expected as generally one male is spawned with one female at the hatchery.

116 Relative reproductive success for parent pairs with one or more offspring, across years 0.8

2006

2007

2008

2009 0.6 0.4 Frequency 0.2 0.0 1 2 3 4 5 6 7 8 9 10 11 12 13

Number of Offspring

Figure 3.4: Number of offspring (full-siblings) that returned to the hatchery for parents spawned in each study year, 2006-2009. Note that offspring of 2009 spawners are under- represented as sampling permitted assignment of only two-year old fish.

117 12 10 8 6 Number of Offspring Number 4 2

700 800 900 1000

Length of Mother (mm)

Figure 3.5: Relationship between the length of a mother and the number of her oﬀspring that returned to the hatchery as adults at ages two, three or four. The size of full-sibling families here ranges from one to thirteen.

118 Table 3.3: Heritability (h2) of length-at-maturity estimated as the slope of the length- length regression line between different comparisons of parents and offspring (Fig- ure 3.6). Mean is the average length of the parents. The regression goodness of fit (R2) and standard error (SE) are also reported.

Parent Mean Mean Mean Male Female Oﬀspring All Male Female Male Female h2 0.189 0.139 0.156 0.062 0.110 R2 0.029 0.020 0.036 0.010 0.026 SE 18.93 24.27 19.47 15.98 15.60 oﬀspring of the same gender as the parent, the mother’s length explained more of the

2 variation in the length of her female oﬀspring (F1,1160 = 31.11, R = 0.026, P < 0.001),

2 than did the father’s length of his male oﬀspring (F1,1139 = 11.2, R = 0.010, P <

0.001). Heritability (h2) was calculated as the slope of the length-length regression line.

Of the comparisons examined here, heritability was highest for the mean parent length as realized by all offspring (h2 = 0.189), followed by the mean parent length and female offspring (h2 = 0.156). The heritability of the mother’s length by her female offspring

(h2 = 0.110) was higher than the heritability of the father’s length by his male oﬀspring

(h2 = 0.062; Table 3.3).

3.3.6 Relatedness

The Rxy estimator was used to calculate relatedness between all pairs of individuals within each of the sample collections. Over all samples, Rxy ranged from -0.62 to 0.83 (mean, 0.003), while the mean of all pairwise Rxy values within each broodstock collection (2006-2011) ranged from -0.0067 in the 2006 spring sample to 0.0074 in the

2008 fall sample. The estimator was normally distributed for all collections and overall,

119 900 800 700 Length of All 3-yr old Offspring (mm) of 3-yr old All Length 600

600 700 800 900 1000

Mean Length of Parents (mm) 950 900 900 850 800 800 750 700 700 Length of Male 3-yr old Offspring (mm) 3-yr old of Male Length Length of Female 3-yr old Offspring (mm) 3-yr old of Female Length 650 600 600 700 800 900 1000 600 700 800 900 1000

Mean Length of Parents (mm) Mean Length of Parents (mm) 950 900 900 850 800 800 750 700 700 Length of 3-yr old Male Offspring (mm) Male of 3-yr old Length Length of 3-yr old Female Offspring (mm) Female of 3-yr old Length 650 600 600 500 600 700 800 900 1000 1100 650 700 750 800 850 900 950 1000

Length of Father (mm) Length of Mother (mm)

Figure 3.6: Linear regression of parental length on the length of their 3-year old adult offspring. Independent comparisons were made for: mean parent length and all offspring, male offspring, and female offspring, as well as, fathers and male offspring and mothers and female offspring.

120 while the skew was low but positive for all but the fall-run 2008 collection (range, -0.017 to 0.068; mean, 0.044; Figure 3.7). A positive skew suggests an asymmetry towards Rxy values greater than zero (i.e. a longer tail of higher relatedness estimates).

The distribution of relatedness between parents that successfully had offspring return to the hatchery as adults was compared to that of parent pairs with no reproductive success for the 2006, 2007, 2008 and 2009 spring-run brood years. These were the four samples for which the mated pair was recorded during spawning at the hatchery, however, only for the years 2006 and 2007 was the full age structure of the cohort (two-, three- and four-year olds) recovered through parentage assignments. It is likely that a small proportion of the parent pairs from 2008 deemed unsuccessful, may yet have four-year old offspring return in 2012, and likewise, many of the unsuccessful parent pairs from 2009 will have three-year old offspring return in 2012 and four-year olds in

2013. Data were again, approximately normally distributed. For all successful parent pairs (across years), Rxy ranged from -0.34 to 0.34 (mean, -0.0083) with a skew of 0.002; for unsuccessful pairs, Rxy ranged from -0.33 to 0.43 (mean, 8×10−4) with a skew of

0.084 (Figure 3.8). For all years 2006-2009, mean relatedness was larger for unsuccessful parent pairs (range, -0.0296 - 0.0039) than for successful parent pairs (range, -0.0178 -

-0.0035). T-tests detected no signiﬁcant diﬀerences (mean p-value = 0.326) in the mean of Rxy values between successful and unsuccessful parent pairs in any year, or overall.

With the exception of 2008, skew was also more positive in the sample of unsuccessful spawners (range, -0.047 - 0.211) than among successful spawners (range, -0.475-0.148).

A weak negative correlation was detected between the relatedness of successful spawn-

121 2006 Spring [N = 955] 2007 Spring [N = 1255] 4 4

3 Mean = -0.0066 3 Mean = -0.0046 Std.Dev = 0.121 Std.Dev = 0.122

2 Range= -0.56 - 0.77 2 Range= -0.62 - 0.8

Density Skew= 0.068 Density Skew= 0.04

1 Histogram density 1 Histogram density Normal distribution Normal distribution 0 0

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Distribution of rxy Distribution of rxy

2008 Spring [N = 961] 2008 Fall [N = 2837] 4 4

3 Mean = 0.0064 3 Mean = 0.0074 Std.Dev = 0.12 Std.Dev = 0.12

2 Range= -0.54 - 0.76 2 Range= -0.58 - 0.74

Density Skew= 0.006 Density Skew= -0.017

1 Histogram density 1 Histogram density Normal distribution Normal distribution 0 0

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Distribution of rxy Distribution of rxy

2009 Spring [N = 806] 2010 Spring [N = 1203] 4 4

3 Mean = 6e-04 3 Mean = 6e-04 Std.Dev = 0.122 Std.Dev = 0.123

2 Range= -0.55 - 0.78 2 Range= -0.58 - 0.83

Density Skew= 0.062 Density Skew= 0.087

1 Histogram density 1 Histogram density Normal distribution Normal distribution 0 0

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Distribution of rxy Distribution of rxy

2011 Spring [N = 1101] All Years [N = 9118] 4 4

3 Mean = 0.0037 3 Mean = 0.0033 Std.Dev = 0.122 Std.Dev = 0.121

2 Range= -0.6 - 0.82 2 Range= -0.62 - 0.83

Density Skew= 0.059 Density Skew= 0.029

1 Histogram density 1 Histogram density Normal distribution Normal distribution 0 0

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Distribution of rxy Distribution of rxy

Figure 3.7: Distribution of the relatedness coeﬀcient (Rxy; Queller and Goodnight

1989) between all possible pairs of individuals in each collection of spawning broodstock and over all samples. Values are normally distributed, so the range, mean, standard deviation (Std. Dev.) and skew are reported.

122 2 ing pairs (all years) and the number of oﬀspring they produced (F1,1651 = 4.011, R =

0.002, P < 0.05), suggesting that less related parents may realize greater reproductive success (Figure 3.9).

3.3.7 Fishery Samples

A total of 24,242 Chinook salmon were sampled in four ﬁshery collections from

2010 to 2012; of these, 874 (3.6%) were excluded for excessive missing data ( >10 loci), leaving 23,368 samples for parentage analysis. A total of 771 ﬁsh sampled in ocean

fisheries were assigned to FRH parents (Table 3.4). Over all assigned fishery samples, mean FDR was 0.001 and the posterior probability of the parent-offspring relationship ranged from 0.879 to 0.999 (mean, 0.996). Of the 2090 samples collected at California ports in 2010, 1855 were successfully genotyped, and CWTs were recovered for 1108

(515 from the Feather River). Recovered CWTs identiﬁed 61 individuals from the FRH spring-run, 40 of which were conﬁrmed by parentage analysis (65.6%). Nine additional individuals that were assigned to spring-run parents presumably had lost or unreadable

CWTs (the hatchery reports 100% tagging of spring-run offspring). One individual which assigned to spring-run parents with high confidence (FDR = 0) contained a CWT indicating a Coleman National Fish Hatchery (located on Battle Creek, a more northern tributary to the Sacramento River) origin, however, the cross between the genetically assigned parents had been recorded at the FRH in 2007, strongly suggesting a misread or errantly placed CWT. Importantly, only two-year old ocean-caught fish in 2010 were available for assignment to the fall-run broodstock, which was sampled only in 2008.

123 Successful pairs 2006 [N = 237] Unsuccessful pairs 2006 [N = 82] 5 5

4 Mean = -0.0035 4 Mean = -0.0296 Std.Dev = 0.114 Std.Dev = 0.131 3 Range= -0.34 - 0.31 3 Range= -0.33 - 0.31

2 Skew= 0.087 2 Skew= 0.211

Density Histogram density Density Histogram density 1 Normal distribution 1 Normal distribution 0 0

-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0